Financial · compiled 3 September 2026 · 63 sources

The economics of AI distillation: prices, training costs, and the market shocks

Distillation is, at bottom, an arbitrage: the capability embedded in a $40M-$500M frontier training run can be harvested through an API for a four- or five-figure query bill and re-trained into a small model for hundreds of dollars. That asymmetry is now visible in published price lists, where the gap between a vendor's flagship and its small tier runs 10x to 42x on output tokens, and in the research record, where Sky-T1-32B was distilled for under $450 and s1-32B for a reported ~$50. It became a macro event on 27 January 2025, when DeepSeek-R1 and its six open-weight distilled students wiped $589B off Nvidia's market capitalisation in a single session - the largest one-day loss in stock-market history - and it became a policy event in February 2026 when OpenAI told the House Select Committee on China that DeepSeek was running obfuscated distillation pipelines against its models. The counter-trend matters too: Epoch AI measures inference prices for a fixed capability level falling 9x-900x per year, which compresses the payback on self-hosting a distilled model to the point where, on our own TCO model, a single-GPU deployment only beats Claude Haiku 4.5 above ~658M output tokens a month and never beats the cheapest serverless open-model endpoints. And in August 2026 DeepSeek reversed the race to zero, raising V4 API prices by as much as 1,100% and introducing peak/off-peak rates - the first major signal that ultra-cheap distilled inference was being priced against capacity, not against marginal cost.

Key figures · 8 figures

Nvidia single-day market-cap loss

589 USD billions

-17% in one session

27 Jan 2025, after DeepSeek-R1 and its distilled students shipped. Largest one-day loss in US market history.

cnbc.com

Cost to train Sky-T1-32B by distillation

450 USD

8x H100 for 19 hours

Matches o1-preview on Math500 and AIME24; teacher was QwQ-32B-Preview, base was Qwen2.5-32B-Instruct.

novasky-ai.github.io

DeepSeek-R1 reinforcement-learning training cost

294,000 USD

512 H800s x 80 hours

Disclosed in the peer-reviewed Nature paper, Sept 2025. Excludes the ~$5.6M base-model run and all R&D, data and infrastructure.

cnn.com

Frontier-vs-small output price spread, OpenAI

42 x

$50 vs $1.20 per MTok

gpt-6-astra output $50/MTok against gpt-5.6-luna at $1.20/MTok on the same list.

developers.openai.com

Inference price decline for fixed capability

900 x per year (upper bound)

range 9x-900x

Epoch AI, measuring the cheapest model clearing a fixed benchmark threshold. GPT-4-level fell from $37.50/MTok (Mar 2023) to $0.18/MTok (Feb 2025).

epoch.ai

Frontier training-run cost growth

2.4 x per year

95% CI 2.0x-3.1x, since 2016

Epoch AI, amortised hardware + energy across 45 frontier models; cloud-rental method gives 2.6x/yr.

epoch.ai

Teacher-query bill to rebuild an 800k-trace reasoning corpus

40,000 USD (upper end)

$1,920 at the cheap end

Author calculation: 800k traces x 2,000 output tokens = 1.6B output tokens, priced at 2026 list rates from Claude Opus 5 ($25/MTok) down to gpt-5.6-luna ($1.20/MTok).

platform.claude.com

DeepSeek V4 API price increase

1,100 % (maximum)

effective 16 Aug 2026

Cache-hit input tokens rose up to 1,100%; output tokens 127%-371%. Peak/off-peak schedule introduced for the first time.

infoworld.com

Key findings · 10 findings

  1. The distillation arbitrage is a five-order-of-magnitude gap, and it is widening

    Epoch AI puts the final training run of GPT-4 at roughly $40M on an amortised-hardware basis (the Stanford AI Index puts it at $78M on cloud-rental accounting), and frontier run costs have grown 2.4x per year since 2016. Against that, Sky-T1-32B was distilled for under $450 of GPU time and TinyZero reproduced R1-Zero-style behaviour for under $30. Even the teacher-query cost is small: at 2026 list prices, regenerating an 800k-trace reasoning corpus like the one behind DeepSeek's R1-Distill family costs $1,920 (gpt-5.6-luna) to $40,000 (Claude Opus 5). The ratio between building the capability and copying it is roughly 1,000:1 to 100,000:1.

    Sources epoch.ai · novasky-ai.github.io · developers.openai.com

  2. Every major vendor now sells its own distillation discount as a product tier

    The frontier-to-small spread on output tokens is 10x at Anthropic (Claude Fable 5.1 at $50/MTok vs Haiku 4.5 at $5/MTok), 42x at OpenAI (gpt-6-astra $50 vs gpt-5.6-luna $1.20), 30x at Google (Gemini 3.1 Pro $12 vs Gemini 2.5 Flash-Lite $0.40) and 75x at Mistral (Medium 3.5 at $7.50 vs Ministral 3 3B at $0.10). Vendors capture the distillation margin internally rather than losing it to third parties. The commercial logic is that a customer who would otherwise self-distill can be retained at a price point the vendor still profits at.

    Sources platform.claude.com · developers.openai.com · ai.google.dev · mistral.ai

  3. Self-hosting a distilled model is now the expensive option for most buyers

    On our TCO model (1x H100 SXM at $2.40/GPU-hour on demand, 24/7, plus 0.25 FTE of MLOps at $200k/yr fully loaded = $5,919/month), a self-hosted distilled model only undercuts Claude Opus 5 above ~132M output tokens a month, Claude Sonnet 5 above ~329M, and Claude Haiku 4.5 above ~658M. It never undercuts gpt-5.6-luna or DeepSeek V4-Flash within the throughput capacity of a single GPU. Serverless open-model endpoints have eaten the economic case for owning inference except at genuine scale or where data residency forces it.

    Sources gmicloud.ai · platform.claude.com · api-docs.deepseek.com

  4. The January 2025 shock repriced compute, not software

    DeepSeek-R1 shipped on 20 January 2025 with six open-weight distilled students under MIT licence. Seven days later Nvidia fell 17% and lost $589B of market capitalisation, the Nasdaq 100 fell 3%, and the semiconductor index had its worst day since March 2020. The market read distillation as a claim that frontier capability could be reproduced without frontier capital expenditure. It was wrong on the timescale: Nvidia crossed $5T in October 2025 - the first company ever to do so - and stood at roughly $5.43T on 2 September 2026.

    Sources cnbc.com · thenationalnews.com · stockanalysis.com

  5. DeepSeek's $5.6M figure was a marginal-cost number, and the dispute is about accounting

    The DeepSeek-V3 technical report states 2.788M H800 GPU-hours for full training; the widely circulated $5.576M is that figure multiplied by an assumed $2/GPU-hour rental rate. SemiAnalysis countered on 31 January 2025 that DeepSeek's server capex is around $1.6B across roughly 50,000 Hopper GPUs with about $944M of operating cost, and that the published number excludes R&D, data, failed runs and hardware total cost of ownership. The Nature paper on R1 (Sept 2025) is narrower still: $294,000 covers only the reinforcement-learning stage on top of an already-trained base.

    Sources arxiv.org · semianalysis.com · theregister.com

  6. The race to zero reversed in August 2026

    DeepSeek warned on 6 August 2026 of a significant price increase and implemented it on 16 August: V4-Flash output went from $0.28/MTok to $0.66 off-peak and $1.32 at peak; V4-Pro output from $0.87 to $1.98/$3.96. Cache-hit input tokens rose by as much as 1,100%. Seventeen of twenty-four hours remain at the half-price off-peak rate, and peak hours are set on Beijing business time, so the increase falls hardest on domestic users and lightest on Western buyers. The signal is that ultra-cheap distilled inference was capacity-constrained, not structurally free.

    Sources infoworld.com · api-docs.deepseek.com · caixinglobal.com

  7. Cloud vendors monetise distillation through fine-tuning and provisioned throughput, not through the distillation itself

    OpenAI's Model Distillation ships Stored Completions free and charges standard fine-tuning rates ($25/MTok training for gpt-4.1, $1.50/MTok for gpt-4.1-nano). Amazon Bedrock Model Distillation (GA 1 May 2025) charges for the teacher inference calls used to synthesise data, then bills the resulting custom model at $1.95/month storage plus Provisioned Throughput - there is no on-demand tier for a distilled model on Bedrock at any volume. AWS markets distilled models as up to 500% faster and up to 75% cheaper to run with under 2% accuracy loss on RAG. The workflow is free; the lock-in is in where the student runs.

    Sources developers.openai.com · docs.aws.amazon.com · press.aboutamazon.com

  8. Distillation allegations moved from a commercial dispute to a congressional one

    Microsoft security researchers observed large-scale data exfiltration through OpenAI developer accounts they linked to DeepSeek in late 2024; the probe became public on 29 January 2025. The House Select Committee on the CCP concluded in its April 2025 report that it is highly likely DeepSeek used unlawful model distillation techniques. In February 2026 OpenAI submitted a memo to the same committee describing sophisticated, multi-stage distillation pipelines using obfuscated third-party routers to conceal origin. The financial stake is that a distillation attack converts a multi-hundred-million-dollar capital asset into a commodity a competitor can rent.

    Sources bloomberg.com · techpolicy.press · cdn.openai.com

  9. Model extraction is cheap enough to be an operating expense, not a capital project

    Carlini et al. recovered the exact hidden dimension of gpt-3.5-turbo and estimated the full embedding-projection matrix could be extracted for under $2,000 in API queries; a limited version of the attack cost under $200, and ada and babbage were fully extracted for under $20. Combined with the corpus-generation figures above, the total cash cost of a serious behavioural-cloning effort against a frontier model sits in the $10^3-$10^5 range against a $10^8 asset. No defensive spend scales down to that.

    Sources arxiv.org · proceedings.mlr.press

  10. Capital followed the small-model thesis, but the exits were modest

    Arcee AI raised a $24M Series A led by Emergence Capital for domain-specific small language models. Predibase, which sold fine-tuning tooling for small open models, raised over $28M and was acquired by Rubrik in June 2025 for a reported $100M-$500M. Together AI, the largest pure-play open-model inference platform, raised $305M at $3.3B in February 2025 and $800M at $8.3B in July 2026. Mistral raised a EUR 1.7B Series C at EUR 11.7B in September 2025 with ASML taking 11%. The value accrued to inference capacity and to European sovereignty plays, not to distillation tooling as a standalone category.

    Sources arcee.ai · techcrunch.com · techcrunch.com · mistral.ai

Charts · 7 charts

Output price per million tokens: frontier tier vs small tier

USD/MTok
The values plotted in “Output price per million tokens: frontier tier vs small tier”, in USD/MTok.
VendorFrontier tier USD/MTokSmall / distilled tier USD/MTok
OpenAI (gpt-6-astra)50
Anthropic (Fable 5.1)50
Anthropic (Opus 5)25
Google (Gemini 3.1 Pro)12
Mistral (Medium 3.5)7.5
xAI (grok-4.6)6
Alibaba (qwen3.8-max)6
DeepSeek (V4-Pro peak)3.96
Meta (Llama 3.3 70B)1.04
OpenAI (gpt-5.6-luna)1.2
Anthropic (Haiku 4.5)5
Google (Gemini 2.5 Flash-Lite)0.4
Mistral (Ministral 3 3B)0.1
xAI (grok-build-0.1)2
Alibaba (qwen-turbo)0.2
DeepSeek (V4-Flash off-peak)0.66
Meta (Llama 3 8B Lite)0.14

Standard list prices as of 3 September 2026. Anthropic appears twice because Opus 5 and Fable 5.1 sit at different points on the same line. Meta prices are Together AI's serverless rates, since Meta does not sell a first-party API for these models.

Sources: developers.openai.com · platform.claude.com · ai.google.dev · mistral.ai · docs.x.ai · alibabacloud.com · api-docs.deepseek.com · together.ai

Price of a fixed capability level, 2021-2026

USD/MTok
The values plotted in “Price of a fixed capability level, 2021-2026”, in USD/MTok.
DateGPT-3 level (MMLU 42) - a16z USD/MTokGPT-3.5 level (MMLU >= 64.8) - Epoch USD/MTokGPT-4 level (MMLU >= 86) - Epoch USD/MTokGPQA Diamond >= 50 - Epoch USD/MTok
2021-1160
2024-110.06
2022-1120
2024-100.07
2023-0337.5
2025-020.18
2023-1115
2024-120.12

Plot on a log y-axis. Each series has only the two endpoints published by the source; the intermediate path was not disclosed as a series. Epoch's aggregate finding across six benchmarks is a 9x-900x annual decline with a median near 50x.

Sources: epoch.ai · a16z.com

Monthly bill vs monthly volume: when does self-hosting a distilled model win?

USD
The values plotted in “Monthly bill vs monthly volume: when does self-hosting a distilled model win?”, in USD.
Output tokens per month (millions)Claude Opus 5 (teacher) USDClaude Sonnet 5 USDClaude Haiku 4.5 (vendor small tier) USDgpt-5.6-luna (vendor nano tier) USDDeepSeek V4-Flash off-peak USDSelf-hosted distilled 8B, 1x H100 + 0.25 FTE USD
145189225,919
104501809020155,919
502,250900450100775,919
1004,5001,8009002001545,919
2009,0003,6001,8004003085,919
50022,5009,0004,5001,0007705,919
1,00045,00018,0009,0002,0001,5405,919

Author-computed. Assumes 4 input tokens per output token. Self-hosted line is flat at $5,919/month ($1,752 GPU + $4,167 staffing) up to ~1,051M output tokens/month capacity. Crossings: 132M vs Opus 5, 329M vs Sonnet 5, 658M vs Haiku 4.5. The gpt-5.6-luna and DeepSeek lines never cross within capacity.

Sources: platform.claude.com · developers.openai.com · api-docs.deepseek.com · gmicloud.ai

Training cost: frontier runs vs distilled students

USD
The values plotted in “Training cost: frontier runs vs distilled students”, in USD.
Model / runUSD (log scale) USD
DeepSeek infrastructure (SemiAnalysis est.)1,600,000,000
Gemini Ultra 1.0 (AI Index)191,000,000
GPT-4 (AI Index)78,000,000
GPT-4 (Epoch amortised)40,000,000
DeepSeek-V3 final run (claimed)5,576,000
DeepSeek-R1 RL stage (Nature)294,000
Sky-T1-32B450
s1-32B50
TinyZero30

Log scale spans eight orders of magnitude. The bars are not accounted on a common basis - see the training-cost table notes. The point of the chart is the shape of the ladder, not a like-for-like comparison.

Sources: epoch.ai · semianalysis.com · arxiv.org · novasky-ai.github.io · cnn.com

Teacher-query bill to assemble a distillation corpus, 2026 list prices

USD
The values plotted in “Teacher-query bill to assemble a distillation corpus, 2026 list prices”, in USD.
Teacher model17k traces (Sky-T1 / Bespoke-Stratos scale) USD800k traces (DeepSeek R1-Distill scale) USD
Claude Opus 585040,000
gpt-5.6-sol68032,000
Gemini 3.5 Flash30614,400
grok-4.62049,600
deepseek-v4-pro (off-peak)673,168
gpt-5.6-luna411,920
gpt-oss-120b (Groq)20960

Author calculation: 2,000 output tokens per trace, input cost ignored. Halve every bar again if the Batch API 50% discount applies.

Sources: platform.claude.com · developers.openai.com · ai.google.dev · api-docs.deepseek.com · console.groq.com

Nvidia market capitalisation around the distillation shocks

USD billions
The values plotted in “Nvidia market capitalisation around the distillation shocks”, in USD billions.
EventMarket-cap changeMarket-cap level
27 Jan 2025: DeepSeek-R1 shock (one session)-589
14 May - 8 Jul 2026: drawdown from peak-1,000
Oct 2025: first $5T company5,060
2 Sep 20265,430

The 27 January 2025 move was a 17% single-session fall and the largest one-day market-cap loss in US stock-market history; the Nasdaq 100 fell 3% and the S&P 500 1.5% the same day. The 2026 drawdown was attributed to rotation into memory and storage semiconductors rather than to distillation news.

Sources: cnbc.com · finance.yahoo.com · stockanalysis.com

DeepSeek V4 pricing before and after 16 August 2026

USD/MTok
The values plotted in “DeepSeek V4 pricing before and after 16 August 2026”, in USD/MTok.
Model and token typeBefore 16 Aug 2026 USD/MTokAfter, off-peak USD/MTokAfter, peak USD/MTok
V4-Flash input (cache miss)0.140.220.44
V4-Flash output0.280.661.32
V4-Pro input (cache miss)0.4350.661.32
V4-Pro output0.871.983.96

Peak hours are 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday - Beijing business hours - so 17 of 24 hours stay at the off-peak rate and Western buyers are largely insulated. Cache-hit input tokens, not shown here, rose by as much as 1,100%.

Sources: api-docs.deepseek.com · infoworld.com

Tables · 8 tables

Frontier vs small-tier list prices, September 2026

11 rows
Frontier vs small-tier list prices, September 2026 — Each vendor's flagship compared with its cheapest general-purpose tier, on the same published price list. Prices are USD per million tokens, standard (non-batch, non-cached) rates. — Units: Frontier in in USD/MTok; Frontier out in USD/MTok; Small in in USD/MTok; Small out in USD/MTok; Output price ratio in x.
VendorFrontier modelFrontier in USD/MTokFrontier out USD/MTokSmall / distilled tierSmall in USD/MTokSmall out USD/MTokOutput price ratio x
OpenAIgpt-6-astra1050gpt-5.6-luna0.21.241.7 developers.openai.com
OpenAIgpt-5.6-sol420gpt-5.4-nano0.21.2516 developers.openai.com
AnthropicClaude Fable 5.11050Claude Haiku 4.51510 platform.claude.com
AnthropicClaude Opus 5525Claude Haiku 4.5155 platform.claude.com
GoogleGemini 3.1 Pro Preview212Gemini 2.5 Flash-Lite0.10.430 ai.google.dev
xAIgrok-4.6 (<200k)26grok-build-0.1123 docs.x.ai
DeepSeekdeepseek-v4-pro (peak)1.323.96deepseek-v4-flash (off-peak)0.220.666 api-docs.deepseek.com
MistralMistral Medium 3.51.57.5Ministral 3 (3B)0.10.175 mistral.ai
Alibaba (Qwen)qwen3.8-max26qwen-turbo0.050.230 alibabacloud.com
Meta (via Together AI)Llama 3.3 70B1.041.04Llama 3 8B Instruct Lite0.140.147.4 together.ai
OpenAI open weights (via Groq)gpt-oss-120b0.150.6gpt-oss-20b0.0750.32 console.groq.com

Ratios are output-token price ratios, computed from the listed figures. DeepSeek peak hours are 01:00-04:00 and 06:00-10:00 UTC Monday-Friday; the other 17 hours are half price. xAI and Anthropic rates shown are for prompts below the long-context threshold.

Sources: developers.openai.com · platform.claude.com · ai.google.dev · docs.x.ai · api-docs.deepseek.com · mistral.ai · alibabacloud.com · together.ai · console.groq.com

What it costs to build a model vs what it costs to copy one

11 rows
What it costs to build a model vs what it costs to copy one — Disclosed and estimated training costs, ordered from frontier pre-training down to hobbyist distillation. Note that these figures are not accounted on a common basis - read the note column. — Units: Reported cost in USD.
Model / runOrganisationReported cost USDComputeDateWhat the number covers
Gemini Ultra 1.0Google DeepMind191,000,000TPU v4, ~35 MW2023-12Stanford AI Index cloud-rental accounting; Epoch's amortised-hardware method gives ~$30M epoch.ai
GPT-4OpenAI78,000,000A100 cluster2023-03Stanford AI Index cloud-rental accounting; Epoch amortised-hardware estimate ~$40M epoch.ai
GPT-4 (amortised)OpenAI40,000,000A100 cluster2023-03Epoch AI: depreciated hardware + energy for the final run only epoch.ai
DeepSeek infrastructure (all-in estimate)DeepSeek / High-Flyer1,600,000,000~50,000 Hopper GPUs2025-01SemiAnalysis estimate of total server capex; ~$944M of operating cost on top semianalysis.com
DeepSeek-V3 (final run, claimed)DeepSeek5,576,0002.788M H800 GPU-hours2024-12GPU-hours are in the technical report; the dollar figure applies an assumed $2/GPU-hour rental rate arxiv.org
DeepSeek-R1 (RL stage)DeepSeek294,000512 H800s x 80 hours (~41k GPU-hours)2025-09Disclosed in the Nature paper; excludes the V3 base model, data, energy, infrastructure and staff cnn.com
DeepSeek-R1-Distill-Qwen-32BDeepSeekSFT on 800k R1-generated samples2025-01Cost undisclosed; the 800k-sample corpus is the expensive input, not the fine-tune arxiv.org
Bespoke-Stratos-32BBespoke Labs8x H100 for 27 hours (216 GPU-hours)2025-01Cost undisclosed. 17k traces distilled from DeepSeek-R1 in 1.5 hours of teacher inference; 47x fewer examples than R1-Distill-Qwen-32B bespokelabs.ai
Sky-T1-32B-PreviewNovaSky, UC Berkeley4508x H100 for 19 hours (152 GPU-hours)2025-01Full disclosed cost of the fine-tune; 17k traces from QwQ-32B-Preview reformatted with GPT-4o-mini novasky-ai.github.io
s1-32BStanford / University of Washington5016x H100 for 26 minutes (~6.9 GPU-hours)2025-02GPU time is stated in the paper; the ~$50 figure comes from press coverage, not the paper itself. 1,000 traces distilled from Gemini Thinking Experimental arxiv.org
TinyZeroUC Berkeley (Jiayi Pan et al.)303B Qwen base, RL on Countdown task2025-01Server cost for the experiments only; reproduces R1-Zero-style self-verification on a narrow task dailycal.org

The costs in this table are NOT comparable on a like-for-like basis. Frontier rows are whole pre-training runs; distillation rows are fine-tunes that assume a free base model and a paid-for teacher. The honest comparison is the ratio between the top of the table and the bottom, which is roughly 10^5 to 10^6.

Sources: epoch.ai · semianalysis.com · arxiv.org · novasky-ai.github.io · arxiv.org

What it costs to buy a teacher's reasoning traces at 2026 list prices

7 rows
What it costs to buy a teacher's reasoning traces at 2026 list prices — Author calculation. Assumes an average reasoning trace of 2,000 output tokens and ignores input-token cost (prompts are short relative to reasoning traces). Two corpus sizes: the 17k traces used by Sky-T1 and Bespoke-Stratos, and the 800k samples DeepSeek used for its R1-Distill family. — Units: Output price in USD/MTok; 17k traces (34M out-tok) in USD; 800k traces (1.6B out-tok) in USD.
Teacher modelOutput price USD/MTok17k traces (34M out-tok) USD800k traces (1.6B out-tok) USD
Claude Opus 52585040,000 platform.claude.com
gpt-5.6-sol2068032,000 developers.openai.com
Gemini 3.5 Flash930614,400 ai.google.dev
grok-4.662049,600 docs.x.ai
deepseek-v4-pro (off-peak)1.98673,168 api-docs.deepseek.com
gpt-5.6-luna1.2411,920 developers.openai.com
gpt-oss-120b (via Groq)0.620960 console.groq.com

Applying the Batch API discount (50% at Anthropic and OpenAI) halves every figure again. These are list prices for legitimate API use; a distillation programme that violates terms of service would face the same compute bill plus the cost of evading detection, which OpenAI's February 2026 memo to Congress describes as obfuscated third-party routers.

Sources: platform.claude.com · developers.openai.com · api-docs.deepseek.com

TCO: teacher API vs distilled API vs self-hosted, by monthly volume

7 rows
TCO: teacher API vs distilled API vs self-hosted, by monthly volume — Author-computed model. Assumptions: 4 input tokens per output token (typical RAG-style workload); monthly cost = output_MTok x (4 x input_price + output_price). Self-hosted row = 1x H100 SXM rented on demand at $2.40/GPU-hour for 730 hours ($1,752/mo) plus 0.25 FTE of MLOps at $200,000/yr fully loaded ($4,167/mo) = $5,919/mo flat, with capacity of ~1,051M output tokens/month at a sustained 400 output tok/s. — Units: Output tokens/month in millions; Claude Opus 5 ($5/$25) in USD/mo; Claude Sonnet 5 ($2/$10) in USD/mo; Claude Haiku 4.5 ($1/$5) in USD/mo; gpt-5.6-luna ($0.20/$1.20) in USD/mo; DeepSeek V4-Flash off-peak ($0.22/$0.66) in USD/mo; Self-hosted distilled 8B, 1x H100 in USD/mo.
Output tokens/month millionsClaude Opus 5 ($5/$25) USD/moClaude Sonnet 5 ($2/$10) USD/moClaude Haiku 4.5 ($1/$5) USD/mogpt-5.6-luna ($0.20/$1.20) USD/moDeepSeek V4-Flash off-peak ($0.22/$0.66) USD/moSelf-hosted distilled 8B, 1x H100 USD/mo
145189225,919 platform.claude.com
104501809020155,919 platform.claude.com
502,250900450100775,919 platform.claude.com
1004,5001,8009002001545,919 platform.claude.com
2009,0003,6001,8004003085,919 platform.claude.com
50022,5009,0004,5001,0007705,919 platform.claude.com
1,00045,00018,0009,0002,0001,5405,919 platform.claude.com

Breakevens for the self-hosted option: 132M output tokens/month vs Opus 5, 329M vs Sonnet 5, 658M vs Haiku 4.5, 2,959M vs gpt-5.6-luna and 3,843M vs DeepSeek V4-Flash off-peak. The last two exceed single-GPU capacity, so at these prices a one-GPU deployment never pays back against the cheapest hosted endpoints. Excludes the one-off distillation cost ($450-$40,000, see other tables), redundancy, and the fact that measured self-hosted cost per output MTok on identical H100 hardware spans $0.21 to $15.25 depending purely on request concurrency.

Sources: platform.claude.com · developers.openai.com · api-docs.deepseek.com · gmicloud.ai · arxiv.org

Cloud vendor distillation products and how they charge

7 rows
Cloud vendor distillation products and how they charge — Managed distillation workflows offered by the major platforms, with the actual revenue mechanism.
VendorProductLaunchedHow it chargesStated benefit
OpenAIModel Distillation (Stored Completions + Evals)2024-10Stored Completions free; standard fine-tuning rates apply ($25/MTok training for gpt-4.1, $5 for gpt-4.1-mini, $1.50 for gpt-4.1-nano, $100/hr for o4-mini)Train smaller cost-efficient models on frontier outputs for a specific task developers.openai.com
Amazon Web ServicesAmazon Bedrock Model Distillation2024-12 preview, GA 2025-05-01Teacher inference charged at on-demand rates when Bedrock synthesises data; custom model storage $1.95/model/month; inference only via Provisioned ThroughputUp to 500% faster and up to 75% less expensive to run, with under 2% accuracy loss on RAG docs.aws.amazon.com
Google CloudVertex AI supervised tuning / distillation2024Per training token; tuned model endpoints billed at 1.5x the base model rateTask-specific tuned Gemini Flash and Flash-Lite students cloud.google.com
MicrosoftAzure OpenAI / Microsoft Foundry stored completions and distillation2024-10Standard Azure fine-tuning and inference rates; distillation requires a minimum of 10 stored completionsTurn production traffic against a large model into a fine-tuning set for a small one learn.microsoft.com
AnthropicNo first-party distillation productn/aTiered model line (Fable/Opus/Sonnet/Haiku) plus Batch API 50% discount and prompt caching down to 0.025x input priceVendor captures the cost-reduction margin internally rather than selling a distillation pipeline platform.claude.com
Together AIServerless open-model inference and fine-tuning2023Per token for serverless; dedicated H100 at $3.99/hr promotional ($5.49 regular), B200 at $8.99/hrHost the distilled student without owning hardware together.ai
Fireworks AIServerless and on-demand GPU deployments2023H100 80GB and H200 141GB at $8.00/hr from 1 Sep 2026 (previously $7.00); B200 $13.00/hr; GB300 $20.00/hr; 1.5x premium for region-restricted deploymentsDedicated capacity for custom and distilled models fireworks.ai

Bedrock's constraint is the commercially interesting one: a distilled model cannot be served on demand, so the customer trades a per-token bill for an hourly Provisioned Throughput commitment, which reverses the economics for low-volume users.

Sources: developers.openai.com · docs.aws.amazon.com · aws.amazon.com · together.ai · fireworks.ai

Price of a fixed capability level over time

5 rows
Price of a fixed capability level over time — Measured price of the cheapest model clearing a fixed benchmark threshold, from Epoch AI and a16z. This is the deflation that distillation both causes and competes with. — Units: Price then in USD/MTok; Price then in USD/MTok; Decline in x.
Capability levelFirst availablePrice then USD/MTokCheapest as ofPrice then USD/MTokDecline x
GPT-3 level (MMLU 42)2021-11602024-110.061,000 a16z.com
GPT-3.5 level (MMLU >= 64.8)2022-11202024-100.07286 epoch.ai
GPT-4 level (MMLU >= 86.0)2023-0337.52025-020.18208 epoch.ai
PhD-level science (GPQA Diamond >= 50)2023-11152024-120.12125 epoch.ai
MMLU 83 (GPT-4 launch level)2023-03302024-110.4862 a16z.com

Epoch's headline range across six benchmarks is a 9x-900x annual decline, with a median around 50x and roughly 200x for models released since 2024. a16z's 'LLMflation' framing is a 10x decline per year and 1,000x over three years. The MMLU 83 start price is a16z's stated ~62x reduction applied to the GPT-4 launch price; treat it as derived rather than directly quoted.

Sources: epoch.ai · a16z.com

Capital raised against the small-model / distillation thesis

8 rows
Capital raised against the small-model / distillation thesis — Funding rounds and exits for companies whose pitch is small, distilled or open models. — Units: Amount in USD millions.
CompanyEventAmount USD millionsValuationDateThesis
Arcee AISeed5.5undisclosed2023-12Domain-specific small language models arcee.ai
Arcee AISeries A (Emergence Capital)24undisclosed2024-07Model merging and Spectrum training to cut SLM training cost venturebeat.com
PredibaseTotal VC raised before exit28undisclosed2024Fine-tuning tooling for small open models (Llama, Mistral) techtarget.com
PredibaseAcquired by Rubrik300reported $100M-$500M range2025-06-25Agentic AI needs cheap task-specific models techcrunch.com
Together AISeries B (General Catalyst, Prosperity7)305$3.3B2025-02-20End-to-end platform for building with 200+ open-source models siliconangle.com
Together AISeries C800$8.3B2026-07-01Neocloud inference capacity for open and distilled models techcrunch.com
Mistral AISeries C (ASML lead, EUR 1.3B of EUR 1.7B)1,900EUR 11.7B post-money (~$14B)2025-09-09European open-weight model family from 3B Ministral up to Large mistral.ai
Mistral AIReported raise in progress3,200reported EUR 20B target2026-06Rumoured EUR 3B round; unconfirmed techcrunch.com

Amounts converted to USD millions where the original is in euros, at approximately 1.13 USD/EUR for the September 2025 round (CNBC reported the valuation as ~$14B). The Predibase acquisition amount is the midpoint of a reported $100M-$500M range and should be treated as an estimate, not a disclosed figure. The Mistral 2026 round is rumoured and unconfirmed.

Sources: arcee.ai · techcrunch.com · techcrunch.com · cnbc.com

The cost-of-theft asymmetry, line by line

14 rows
The cost-of-theft asymmetry, line by line — Every published figure that bears on the question: what does it cost to build frontier capability, and what does it cost to take it? — Units: Cost in USD.
Line itemSideCost USDBasis
Gemini Ultra 1.0 final training runBuild191,000,000Stanford AI Index cloud-rental accounting epoch.ai
GPT-4 final training runBuild78,000,000Stanford AI Index cloud-rental accounting epoch.ai
GPT-4 final training run (amortised)Build40,000,000Epoch AI hardware depreciation + energy epoch.ai
DeepSeek-V3 pre-training (claimed)Build5,576,0002.788M H800 GPU-hours at an assumed $2/hr arxiv.org
DeepSeek-R1 RL stageBuild294,000Nature paper: 512 H800s x 80 hours cnn.com
800k-trace reasoning corpus from Claude Opus 5Copy40,000Author calculation at $25/MTok output, 2,000 tokens/trace platform.claude.com
800k-trace reasoning corpus from gpt-5.6-lunaCopy1,920Author calculation at $1.20/MTok output developers.openai.com
Full projection-matrix extraction of gpt-3.5-turbo (estimated)Copy2,000Carlini et al., estimated query cost arxiv.org
Partial model-stealing attack on gpt-3.5Copy200Carlini et al., executed attack arxiv.org
Bespoke-Stratos 17k-trace corpus generationCopy1.5 hours of DeepSeek-R1 inference; dollar cost undisclosed bespokelabs.ai
Sky-T1-32B student fine-tuneCopy4508x H100 for 19 hours novasky-ai.github.io
s1-32B student fine-tuneCopy5016x H100 for 26 minutes; dollar figure from press coverage arxiv.org
Full projection-matrix extraction of ada and babbageCopy20Carlini et al., executed attack arxiv.org
TinyZero R1-Zero-style reproductionCopy30Server cost for the experiments, narrow task only dailycal.org

The build side and copy side are not substitutes: a distilled student inherits behaviour on the distribution it was distilled over, not the teacher's full capability surface. But for the specific task a buyer cares about, the copy side is 3-6 orders of magnitude cheaper, and no legal or technical defence currently scales down to that price point.

Sources: epoch.ai · arxiv.org · novasky-ai.github.io

Timeline · 30 events

  1. product

    OpenAI ships Model Distillation in the API

    Stored Completions (free) plus Evals give developers a managed pipeline to capture frontier outputs and fine-tune a smaller student. Distillation becomes a supported product rather than a research technique.

    Source: openai.com
  2. market

    a16z publishes 'LLMflation'

    Argues inference cost for a fixed capability falls ~10x per year and has fallen 1,000x in three years, from $60/MTok for GPT-3 in Nov 2021 to $0.06/MTok for Llama 3.2 3B on Together.ai.

    Source: a16z.com
  3. product

    AWS announces Amazon Bedrock Model Distillation at re:Invent

    Claims distilled models can be up to 500% faster and up to 75% less expensive to run, with under 2% accuracy loss for RAG use cases.

    Source: press.aboutamazon.com
  4. research

    DeepSeek-V3 technical report discloses 2.788M H800 GPU-hours

    The report gives GPU-hours, not dollars; the widely quoted $5.576M figure is that number multiplied by an assumed $2/GPU-hour rental rate.

    Source: arxiv.org
  5. research

    Sky-T1-32B-Preview trained for under $450

    NovaSky at UC Berkeley distils QwQ-32B-Preview traces into Qwen2.5-32B-Instruct using 17k examples and 19 hours on 8 H100s, matching o1-preview on Math500 (82.4 vs 81.4) and AIME24 (43.3 vs 40.0).

    Source: novasky-ai.github.io
  6. product

    DeepSeek releases R1 plus six open-weight distilled students under MIT licence

    R1-Distill-Qwen-32B scores 72.6 on AIME 2024 and 94.3 on MATH-500; R1-Distill-Llama-70B scores 86.7 on AIME 2024. Frontier-class reasoning becomes free to download.

    Source: huggingface.co
  7. research

    Bespoke-Stratos-32B distilled from DeepSeek-R1 on 17k traces

    Trained on 8xH100 for 27 hours using 47x fewer examples than R1-Distill-Qwen-32B; the R1 traces took 1.5 hours to generate. GPT-4o-mini filtering raised retained-correct-solution rate from 25% to 73%.

    Source: bespokelabs.ai
  8. research

    TinyZero reproduces R1-Zero-style behaviour for under $30

    UC Berkeley graduate researchers use RL on a 3B Qwen base for Countdown and multiplication tasks; the $30 is server cost for the experiments.

    Source: dailycal.org
  9. market

    Nvidia loses $589B of market cap in one session

    Shares fall ~17% from an open of $142.02 to close at $118.50. Largest single-day market-cap loss in US stock-market history. Nasdaq 100 -3%, S&P 500 -1.5%, semiconductor index worst day since March 2020.

    Source: cnbc.com
  10. market

    SemiAnalysis disputes the $5.6M figure

    Estimates DeepSeek's server capex at ~$1.6B across roughly 50,000 Hopper GPUs with ~$944M of operating cost, arguing the published figure covers only GPU time for the pre-training run.

    Source: semianalysis.com
  11. research

    s1-32B: 1,000 traces, 26 minutes, reported ~$50

    Distilled from Gemini Thinking Experimental into Qwen2.5-32B-Instruct on 16 H100s. Exceeds o1-preview on competition maths by up to 27% with budget forcing at inference time.

    Source: arxiv.org
  12. market

    Together AI raises $305M Series B at $3.3B

    General Catalyst and Prosperity7 lead; Nvidia, Salesforce Ventures, Kleiner Perkins participate. The capital funds Blackwell capacity for serving open and distilled models.

    Source: siliconangle.com
  13. policy

    House Select Committee publishes 'DeepSeek Unmasked'

    Concludes it is highly likely DeepSeek used unlawful model distillation techniques against US models, and that the scale was such that V3 often self-identifies as ChatGPT.

    Source: techpolicy.press
  14. product

    Amazon Bedrock Model Distillation reaches general availability

    Distilled models can only be served on Provisioned Throughput, not on demand, which changes the economics for low-volume buyers.

    Source: docs.aws.amazon.com
  15. market

    OpenAI cuts o3 pricing by ~80%

    From $10/$40 to $2/$8 per million input/output tokens, attributed to inference-stack optimisation on the same model. o3-pro launches alongside.

    Source: venturebeat.com
  16. market

    Rubrik acquires Predibase

    Reported at $100M-$500M. Predibase sold fine-tuning tooling for small open models and had raised over $28M from Felicis, Greylock and Sancus Ventures.

    Source: techcrunch.com
  17. product

    OpenAI releases gpt-oss-120b and gpt-oss-20b under Apache 2.0

    First open-weight OpenAI models since GPT-2. gpt-oss-120b has 116.8B total / 5.1B active parameters and matches or exceeds o4-mini on competition coding. Now served at $0.15/$0.60 and $0.075/$0.30 per MTok on Groq.

    Source: openai.com
  18. market

    Mistral raises EUR 1.7B Series C at EUR 11.7B, ASML leads

    ASML puts in EUR 1.3B for ~11% fully diluted. Valuation roughly doubles from EUR 5.8B. Mistral's line spans Ministral 3B at $0.10/$0.10 up to Medium 3.5 at $1.50/$7.50.

    Source: mistral.ai
  19. research

    Nature publishes the DeepSeek-R1 paper with a $294,000 cost figure

    512 H800s for 80 hours for the reinforcement-learning stage. The figure excludes the ~$6M base model, data, energy, infrastructure and staff; commentators including The Register noted the number is not a total cost.

    Source: cnn.com
  20. market

    Nvidia becomes the first $5 trillion company

    Closes at a $5.06T market capitalisation. DeepSeek launches V4-Pro the same day; Nvidia's valuation is unaffected.

    Source: thenationalnews.com
  21. market

    One year on, DeepSeek no longer moves markets

    CNBC reports that the companies hit by the January 2025 selloff have not just recovered but grown, and that subsequent DeepSeek releases produced no comparable investor reaction.

    Source: cnbc.com
  22. policy

    OpenAI memo to the House Select Committee on China

    Accuses DeepSeek of sophisticated multi-stage distillation pipelines using obfuscated third-party routers and unauthorised resellers to conceal origin and evade access restrictions.

    Source: cdn.openai.com
  23. policy

    Frontier Model Forum publishes an issue brief on adversarial distillation

    Frames the risk in terms of how many teacher outputs an attacker can obtain and how much compute they have, but publishes no cost figures.

    Source: frontiermodelforum.org
  24. market

    Together AI raises $800M at $8.3B

    The largest pure-play open-model inference platform roughly 2.5x its valuation in 17 months, on demand for serving open-weight and distilled models.

    Source: techcrunch.com
  25. market

    Nvidia sheds roughly $1T from its 14 May 2026 peak

    Down ~16% from the peak on rotation into memory and storage semiconductors, not on distillation news. Nvidia still holds ~97% of the server GPU market and trades at 18x forward earnings.

    Source: finance.yahoo.com
  26. product

    DeepSeek V4-Flash launches at $0.14/$0.28 per MTok

    Described in the press as accelerating the AI industry's race to zero. The price lasts sixteen days.

    Source: axios.com
  27. market

    DeepSeek warns of a significant API price increase

    Bloomberg and SCMP report the warning ahead of the change, citing surging demand for low-cost models and strained capacity.

    Source: scmp.com
  28. market

    DeepSeek raises V4 prices by up to 1,100% and introduces peak/off-peak rates

    V4-Flash output goes from $0.28 to $0.66 off-peak / $1.32 peak; V4-Pro output from $0.87 to $1.98/$3.96. Cache-hit input tokens rise most. Peak hours are 01:00-04:00 and 06:00-10:00 UTC.

    Source: infoworld.com
  29. market

    Nvidia market capitalisation stands at ~$5.43T

    Up ~28% year on year, and about 9x the value erased on 27 January 2025. The distillation shock repriced sentiment, not the compute build-out.

    Source: stockanalysis.com

Glossary · 16 terms

MTok
One million tokens. The standard unit for LLM API pricing. Input and output tokens are billed at different rates, with output typically 4-5x more expensive.
Blended cost per MTok
A single price figure for a workload with a known input:output ratio. For a 4:1 ratio, blended cost per million output tokens = 4 x input price + output price. Used throughout the TCO tables here.
Teacher API cost
The money paid to a frontier vendor for the inference calls that generate a distillation corpus. Usually the dominant cash cost of a distillation project, larger than the student fine-tune itself.
Amortised training cost
Epoch AI's accounting method: hardware depreciation plus energy over the final training run. Produces figures roughly half those of cloud-rental accounting, which charges the full market rate for every chip-hour.
Cloud-rental accounting
Costing a training run at the price of renting equivalent GPUs on the open market. Used by the Stanford AI Index. Gives $78M for GPT-4 where Epoch's amortised method gives ~$40M.
Marginal training cost
The GPU-hour bill for a single successful run, excluding R&D, failed runs, data acquisition, staff and infrastructure. DeepSeek's $5.6M and $294,000 figures are both marginal costs.
Provisioned Throughput
AWS Bedrock's committed-capacity billing mode, charged hourly per Model Unit. Custom and distilled models on Bedrock can only be served this way, which imposes a fixed monthly floor regardless of volume.
Prompt caching
Charging a reduced rate for repeated prompt prefixes. Anthropic charges 1.25x base input to write a 5-minute cache and 0.1x to read it (0.025x on Fable 5.1 and Mythos 5.1), which can cut effective input cost by up to 90%.
Batch API discount
A 50% reduction on both input and output tokens for asynchronous processing, offered by both OpenAI and Anthropic. Applies to distillation-corpus generation, halving the teacher-query bill.
Peak/off-peak pricing
Time-of-day API pricing, introduced by DeepSeek on 16 August 2026. Peak hours (01:00-04:00 and 06:00-10:00 UTC) cost double the off-peak rate; 17 of 24 hours remain off-peak.
LLMflation
a16z's term for the ~10x-per-year fall in inference cost at a fixed capability level. Epoch AI measures the same phenomenon at 9x-900x per year across six benchmarks, median near 50x.
Breakeven volume
The monthly token volume at which a fixed-cost self-hosted deployment becomes cheaper than a per-token API. Rises as API prices fall, which is why the case for owning inference has weakened even as distillation has got easier.
Cost-of-theft asymmetry
The gap between what it costs to create a capability and what it costs to copy it. For frontier LLMs the ratio is roughly 10^3 to 10^5, because behavioural cloning requires only API access and modest fine-tuning compute.
Model extraction attack
Recovering structural parameters of a black-box model through API queries alone. Carlini et al. recovered the full projection matrix of ada and babbage for under $20 and estimated under $2,000 for gpt-3.5-turbo.
Neocloud
A GPU-specialist cloud provider (Together AI, Fireworks, CoreWeave, RunPod, Vast.ai) that undercuts hyperscalers on GPU-hour pricing. H100 on-demand rates run ~$1.73-$4.00/hr at neoclouds vs $4.00-$8.00 at AWS, GCP and Azure.
Student / teacher
In distillation, the small model being trained (student) and the large model whose outputs it learns from (teacher). Financially: the student is the asset you own, the teacher is the line item on your API bill.

Sources · 63 sources

Every figure on this page comes from one of these primary sources. Compiled 3 September 2026.

  1. OpenAI API PricingOpenAI · 3 September 2026 · pricing
  2. Anthropic Claude PricingAnthropic · 3 September 2026 · pricing
  3. Gemini API PricingGoogle · 3 September 2026 · pricing
  4. DeepSeek API PricingDeepSeek · 3 September 2026 · pricing
  5. Mistral API PricingMistral AI · 3 September 2026 · pricing
  6. xAI Models and PricingxAI · 3 September 2026 · pricing
  7. Alibaba Cloud Model Studio model pricingAlibaba Cloud · 3 September 2026 · pricing
  8. Together AI PricingTogether AI · 3 September 2026 · pricing
  9. Fireworks AI PricingFireworks AI · 3 September 2026 · pricing
  10. Groq supported modelsGroq · 3 September 2026 · pricing
  11. Amazon Bedrock PricingAmazon Web Services · 3 September 2026 · pricing
  12. Customize a model with distillation in Amazon BedrockAmazon Web Services · 2026 · docs
  13. AWS Strengthens Amazon Bedrock with Industry-First AI Safeguard, New Agent Capability and Model CustomizationAmazon · 3 December 2024 · news
  14. Model Distillation in the APIOpenAI · 1 October 2024 · blog
  15. How to use stored completions and distillation in Azure OpenAIMicrosoft · 2026 · docs
  16. Vertex AI pricingGoogle Cloud · 2026 · pricing
  17. Sky-T1: Train your own O1 preview model within $450NovaSky, UC Berkeley · 10 January 2025 · blog
  18. Researchers open source Sky-T1, a reasoning AI model that can be trained for less than $450TechCrunch · 11 January 2025 · news
  19. s1: Simple test-time scalingarXiv (Muennighoff et al.) · 1 February 2025 · paper
  20. Bespoke-Stratos: The unreasonable effectiveness of reasoning distillationBespoke Labs · 22 January 2025 · blog
  21. Campus researchers replicate disruptive Chinese AI for $30The Daily Californian · 31 January 2025 · news
  22. DeepSeek-V3 Technical ReportarXiv (DeepSeek-AI) · 26 December 2024 · paper
  23. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement LearningarXiv (DeepSeek-AI) · 22 January 2025 · paper
  24. DeepSeek-R1-Distill-Llama-70B model cardHugging Face · 20 January 2025 · docs
  25. China's DeepSeek shook the tech world. Its developer just revealed the costCNN · 19 September 2025 · news
  26. DeepSeek didn't really train its flagship model for $294,000The Register · 19 September 2025 · news
  27. DeepSeek Debates: Chinese Leadership On Cost, True Training CostSemiAnalysis · 31 January 2025 · blog
  28. Nvidia sheds almost $600 billion in market cap, biggest drop everCNBC · 27 January 2025 · news
  29. Nvidia stock plummets, loses record $589 billion as DeepSeek prompts questions over AI spendingYahoo Finance · 27 January 2025 · news
  30. Nvidia market cap falls by $1 trillion as stock swoonsYahoo Finance · 8 July 2026 · news
  31. NVIDIA (NVDA) Market Cap & Net WorthStockAnalysis · 2 September 2026 · filing
  32. Will DeepSeek's new AI model crash Nvidia's $5tn party?The National · 25 April 2026 · news
  33. Why DeepSeek didn't cause an investor frenzy again in 2025CNBC · 6 January 2026 · news
  34. Microsoft Probing If DeepSeek-Linked Group Improperly Obtained OpenAI DataBloomberg · 29 January 2025 · news
  35. US House Select Committee Report Accuses DeepSeek of Spying and Circumventing Export Controls on ChipsTech Policy Press · 17 April 2025 · news
  36. OpenAI update to the US House Select Committee on the CCPOpenAI · 12 February 2026 · filing
  37. Issue Brief: Adversarial DistillationFrontier Model Forum · 23 February 2026 · blog
  38. LLM inference prices have fallen rapidly but unequally across tasksEpoch AI · 2025 · blog
  39. How much does it cost to train frontier AI models?Epoch AI · June 2024 · paper
  40. Welcome to LLMflation - LLM inference cost is going down fastAndreessen Horowitz · November 2024 · blog
  41. Stealing Part of a Production Language ModelarXiv (Carlini et al.) · 11 March 2024 · paper
  42. Stealing part of a production language model (ICML 2024)PMLR · 2024 · paper
  43. OpenAI announces 80% price drop for o3VentureBeat · 10 June 2025 · news
  44. gpt-oss-120b & gpt-oss-20b Model CardOpenAI · 5 August 2025 · docs
  45. DeepSeek raises some V4 prices by more than 10x as AI demand strains capacityInfoWorld · 14 August 2026 · news
  46. Tech Brief: DeepSeek Launches V4-Pro and Raises API Prices by as Much as 1,100%Caixin Global · 14 August 2026 · news
  47. DeepSeek signals significant price hike amid surge in demand for low-cost AI modelsSouth China Morning Post · 6 August 2026 · news
  48. DeepSeek's new bargain model accelerates AI's race to zeroAxios · 1 August 2026 · news
  49. Arcee AI secures $24M Series A to transform the landscape of small language modelsArcee AI · July 2024 · blog
  50. Small language models rising as Arcee AI lands $24M Series AVentureBeat · July 2024 · news
  51. Rubrik acquires Predibase to accelerate adoption of AI agentsTechCrunch · 25 June 2025 · news
  52. Rubrik pivots to generative AI with Predibase acquisitionTechTarget · 26 June 2025 · news
  53. Together AI raises $305M for its AI-optimized public cloudSiliconANGLE · 20 February 2025 · news
  54. Neocloud Together AI raises $800M, leaps to $8.3B valuationTechCrunch · 1 July 2026 · news
  55. Mistral AI raises EUR 1.7B to accelerate technological progress with AIMistral AI · 9 September 2025 · blog
  56. AI firm Mistral valued at $14 billion as chip giant ASML takes major stakeCNBC · 9 September 2025 · news
  57. Mistral is rumored to be raising EUR 3B at EUR 20B valuationTechCrunch · 12 June 2026 · news
  58. NVIDIA H100 GPU Pricing: 2026 Rent vs. Buy Cost AnalysisGMI Cloud · 2026 · pricing
  59. H100 Cloud Pricing: Compare 53+ Providers (2026)GetDeploying · 3 September 2026 · pricing
  60. Beyond Per-Token Pricing: A Concurrency-Aware Methodology for LLM Serving CostarXiv · June 2026 · paper
  61. State of AI: An Empirical 100 Trillion Token Study with OpenRouterarXiv / OpenRouter · January 2026 · paper
  62. OpenRouter State of AIOpenRouter · January 2026 · blog
  63. Small Language Model Market Report 2026Research and Markets · 2026 · news