A public compendium · updated daily
How frontier intelligence is compressed, priced, and contested.
Distillation moves capability from a large teacher model into a small, cheap student. It is the quiet engine behind most models people actually pay for, the subject of an open dispute between the largest labs, and a live regulatory question in three jurisdictions. This compendium tracks all of it.
Compare · 74 models · 63 listed prices
Compare teachers and their distilled students
Pick two to four models and the interactive builder puts them side by side on quality, price, latency, context and licence. Every specification it uses is in the tables below, so the comparison can be made by hand, or downloaded as JSON, without running the page.
Every model in the comparison set
74 rows| Model | Vendor | Role | Distilled | Teacher | Input price USD per million tokens | Output price USD per million tokens | GPQA % | Time to first token ms | Context thousand tokens | Licence |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-6 Astra | OpenAI | teacher | no | n/a | 10 | 50 | — | — | 1,050 | Proprietary API |
| GPT-5.6 Sol | OpenAI | teacher | no | n/a | 4 | 20 | 92.4 | — | 1,050 | Proprietary API |
| GPT-5.6 Terra | OpenAI | small-sibling | no | undisclosed | 2 | 12 | 88.4 | — | 1,050 | Proprietary API |
| GPT-5.6 Luna | OpenAI | small-sibling | no | undisclosed | 0.2 | 1.2 | 87 | 1,700 | 1,050 | Proprietary API |
| GPT-5.5 | OpenAI | teacher | no | n/a | 5 | 30 | — | — | — | Proprietary API |
| GPT-5.4 | OpenAI | teacher | no | n/a | 2.5 | 15 | 93 | — | 1,050 | Proprietary API |
| GPT-5.4 mini | OpenAI | small-sibling | no | undisclosed | 0.75 | 4.5 | 88 | — | 400 | Proprietary API |
| GPT-5.4 nano | OpenAI | small-sibling | no | undisclosed | 0.2 | 1.25 | 82.8 | — | 400 | Proprietary API |
| GPT-5 | OpenAI | teacher | no | n/a | 1.25 | 10 | — | — | 400 | Proprietary API |
| GPT-5 mini | OpenAI | small-sibling | no | undisclosed | 0.25 | 2 | 80.3 | — | 400 | Proprietary API |
| GPT-5 nano | OpenAI | small-sibling | no | undisclosed | 0.05 | 0.4 | 70.9 | — | 400 | Proprietary API |
| GPT-4o | OpenAI | teacher | no | n/a | 2.5 | 10 | — | — | 128 | Proprietary API |
| GPT-4o mini | OpenAI | small-sibling | no | undisclosed | 0.15 | 0.6 | — | — | 128 | Proprietary API |
| o3 | OpenAI | teacher | no | n/a | 2 | 8 | 83.3 | — | 200 | Proprietary API |
| o4-mini | OpenAI | small-sibling | no | undisclosed | 1.1 | 4.4 | 81.4 | — | 200 | Proprietary API |
| gpt-oss-120b | OpenAI (open weights) | open | no | undisclosed | 0.15 | 0.6 | — | 850 | 128 | Apache 2.0 |
| gpt-oss-20b | OpenAI (open weights) | open | no | undisclosed | 0.075 | 0.3 | 58.6 | — | 128 | Apache 2.0 |
| Claude Fable 5.1 | Anthropic | teacher | no | n/a | 10 | 50 | — | — | 1,000 | Proprietary API |
| Claude Opus 5 | Anthropic | teacher | no | n/a | 5 | 25 | — | 77,250 | 1,000 | Proprietary API |
| Claude Sonnet 5 | Anthropic | small-sibling | no | undisclosed | 2 | 10 | — | 177,770 | 1,000 | Proprietary API |
| Claude Sonnet 4.6 | Anthropic | small-sibling | no | undisclosed | 3 | 15 | — | — | 1,000 | Proprietary API |
| Claude Haiku 4.5 | Anthropic | small-sibling | no | undisclosed | 1 | 5 | — | 19,920 | 200 | Proprietary API |
| Claude Haiku 3.5 | Anthropic | small-sibling | no | undisclosed | 0.8 | 4 | — | — | 200 | Proprietary API |
| Gemini 3.1 Pro (Preview) | teacher | no | n/a | 2 | 12 | 94.3 | — | 1,000 | Proprietary API | |
| Gemini 3.8 Flash | small-sibling | no | undisclosed | 0.75 | 3.75 | — | — | 1,000 | Proprietary API | |
| Gemini 3.5 Flash | small-sibling | no | undisclosed | 1.5 | 9 | — | — | 1,000 | Proprietary API | |
| Gemini 3.5 Flash-Lite | small-sibling | no | undisclosed | 0.3 | 2.5 | — | 6,480 | 1,000 | Proprietary API | |
| Gemini 3.1 Flash-Lite | small-sibling | no | undisclosed | 0.25 | 1.5 | 72.2 | — | 1,000 | Proprietary API | |
| Gemini 2.5 Pro | teacher | no | n/a | 1.25 | 10 | 86.4 | — | 1,000 | Proprietary API | |
| Gemini 2.5 Flash | distilled | yes | Gemini 2.5 Pro (k-sparse logit distillation) | 0.3 | 2.5 | 82.8 | — | 1,000 | Proprietary API | |
| Gemini 2.5 Flash-Lite | distilled | yes | Gemini 2.5 Pro (k-sparse logit distillation) | 0.1 | 0.4 | — | 300 | 1,000 | Proprietary API | |
| Gemma 4 31B | open | no | undisclosed | — | — | 84.3 | — | 256 | Apache 2.0 | |
| Gemma 4 26B A4B (MoE) | open | no | undisclosed | — | — | 82.3 | — | 256 | Apache 2.0 | |
| Gemma 4 12B Unified | open | no | undisclosed | — | — | 78.8 | — | 256 | Apache 2.0 | |
| Gemma 4 E4B | open | no | undisclosed | — | — | 58.6 | — | 128 | Apache 2.0 | |
| Gemma 4 E2B | open | no | undisclosed | — | — | 43.4 | — | 128 | Apache 2.0 | |
| Gemma 3 27B IT | open | no | undisclosed | — | — | 24.3 | — | 128 | Gemma Terms of Use (custom) | |
| Gemma 3 4B IT | open | no | undisclosed | — | — | 15 | — | 128 | Gemma Terms of Use (custom) | |
| Llama 4 Maverick | Meta | distilled | yes | Llama 4 Behemoth (codistillation) | — | — | 69.8 | 920 | 1,000 | Llama 4 Community License (custom commercial) |
| Llama 4 Scout | Meta | distilled | yes | Llama 4 Behemoth (codistillation) | — | — | 57.2 | — | 10,000 | Llama 4 Community License (custom commercial) |
| Llama 3.3 70B Instruct | Meta | open | no | undisclosed | 1.04 | 1.04 | 50.5 | — | 128 | Llama 3.3 Community License (custom commercial) |
| Llama 3.2 3B Instruct | Meta | distilled | yes | Llama 3.1 8B and 70B (token-level logit distillation after pruning) | — | — | 32.8 | — | 128 | Llama 3.2 Community License (custom commercial) |
| Llama 3.2 1B Instruct | Meta | distilled | yes | Llama 3.1 8B and 70B (token-level logit distillation after pruning) | — | — | 27.2 | — | 128 | Llama 3.2 Community License (custom commercial) |
| DeepSeek-V4-Pro | DeepSeek | teacher | no | n/a | 1.32 | 3.96 | 90.1 | 1,900 | 1,000 | MIT |
| DeepSeek-V4-Flash | DeepSeek | distilled | yes | DeepSeek V4 domain experts (on-policy distillation consolidation) | 0.44 | 1.32 | 88.1 | 1,190 | 1,000 | MIT |
| DeepSeek-V3.2 | DeepSeek | open | yes | DeepSeek specialist models (specialist distillation into the generalist) | — | — | 82.4 | — | 128 | MIT |
| DeepSeek-R1 (0528) | DeepSeek | teacher | no | n/a | — | — | 81 | — | 128 | MIT |
| DeepSeek-R1-Distill-Llama-70B | DeepSeek / Meta base | distilled | yes | DeepSeek-R1 | — | — | 65.2 | — | 128 | MIT (weights) over Llama 3.3 Community License base |
| DeepSeek-R1-Distill-Qwen-32B | DeepSeek / Qwen base | distilled | yes | DeepSeek-R1 | — | — | 62.1 | — | 128 | MIT (weights), Qwen2.5-32B base under Apache 2.0 |
| DeepSeek-R1-Distill-Qwen-14B | DeepSeek / Qwen base | distilled | yes | DeepSeek-R1 | — | — | 59.1 | — | 128 | MIT (weights), Qwen2.5-14B base under Apache 2.0 |
| DeepSeek-R1-Distill-Llama-8B | DeepSeek / Meta base | distilled | yes | DeepSeek-R1 | — | — | 49 | — | 128 | MIT (weights) over Llama 3.1 Community License base |
| DeepSeek-R1-Distill-Qwen-7B | DeepSeek / Qwen base | distilled | yes | DeepSeek-R1 | — | — | 49.1 | — | 128 | MIT (weights), Qwen2.5-Math-7B base under Apache 2.0 |
| DeepSeek-R1-Distill-Qwen-1.5B | DeepSeek / Qwen base | distilled | yes | DeepSeek-R1 | — | — | 33.8 | — | 128 | MIT (weights), Qwen2.5-Math-1.5B base under Apache 2.0 |
| Qwen3.8-27B | Alibaba | open | no | undisclosed | 0.5 | 3 | 89.2 | — | 262 | Apache 2.0 |
| qwen3.8-max | Alibaba | teacher | no | n/a | 2 | 6 | — | — | — | Proprietary API |
| qwen3.8-flash | Alibaba | small-sibling | no | undisclosed | 0.15 | 0.47 | — | — | — | Proprietary API |
| qwen-turbo | Alibaba | small-sibling | no | undisclosed | 0.05 | 0.2 | — | — | — | Proprietary API |
| Qwen3-4B-Instruct-2507 | Alibaba | distilled | yes | Qwen3-32B / Qwen3-235B-A22B (off-policy + on-policy strong-to-weak distillation) | — | — | 62 | — | 262 | Apache 2.0 |
| qwen3-8b (hosted) | Alibaba | distilled | yes | Qwen3-32B / Qwen3-235B-A22B (strong-to-weak distillation) | 0.18 | 0.7 | — | — | 128 | Apache 2.0 |
| Phi-4 (14B) | Microsoft | open | no | undisclosed | — | — | 56.1 | — | 16 | MIT |
| Phi-4-mini-instruct (3.8B) | Microsoft | open | no | undisclosed | — | — | 25.2 | — | 128 | MIT |
| Mistral Medium 3.5 | Mistral AI | teacher | no | n/a | 1.5 | 7.5 | — | — | — | Modified MIT |
| Mistral Large 3 | Mistral AI | open | no | undisclosed | 0.5 | 1.5 | — | — | — | Apache 2.0 |
| Mistral Small 4 | Mistral AI | small-sibling | no | undisclosed | 0.15 | 0.6 | 71.2 | — | — | Apache 2.0 |
| Ministral 3 14B Instruct | Mistral AI | open | no | undisclosed | 0.2 | 0.2 | 71.2 | — | 256 | Apache 2.0 |
| Ministral 3 8B | Mistral AI | open | no | undisclosed | 0.15 | 0.15 | — | — | 256 | Apache 2.0 |
| Ministral 3 3B | Mistral AI | open | no | undisclosed | 0.1 | 0.1 | — | — | 256 | Apache 2.0 |
| Amazon Nova Premier | Amazon | teacher | no | n/a | — | — | — | — | 1,000 | Proprietary API |
| Amazon Nova Pro | Amazon | teacher | no | n/a | 0.8 | 3.2 | 46.9 | — | 300 | Proprietary API |
| Amazon Nova Lite | Amazon | distilled | no | n/a | 0.06 | 0.24 | 42 | — | 300 | Proprietary API |
| Amazon Nova Micro | Amazon | distilled | no | n/a | 0.035 | 0.14 | 40 | — | 128 | Proprietary API |
| Amazon Nova 2 Lite | Amazon | small-sibling | no | undisclosed | 0.3 | 2.5 | — | — | 1,000 | Proprietary API |
| SmolLM3-3B | Hugging Face | open | no | n/a | — | — | 41.7 | — | 128 | Apache 2.0 |
| grok-4.6 | xAI | teacher | no | n/a | 2 | 6 | — | — | 200 | Proprietary API |
List prices for the standard tier on the date each row was compiled; batch and cache discounts excluded.
Published token prices by tier
63 rows| Model | Vendor | Tier | Input price USD per million tokens | Output price USD per million tokens | Parameters billions | Released |
|---|---|---|---|---|---|---|
| gpt-6-astra | OpenAI | frontier | 10 | 50 | — | — |
| gpt-5.6-sol | OpenAI | frontier | 4 | 20 | — | 2026-07 |
| gpt-5.6-terra | OpenAI | frontier | 2 | 12 | — | 2026-07 |
| gpt-5.6-luna | OpenAI | distilled | 0.2 | 1.2 | — | 2026-07 |
| gpt-5.5 | OpenAI | frontier | 5 | 30 | — | — |
| gpt-5.5-pro | OpenAI | frontier | 30 | 180 | — | — |
| gpt-5.4 | OpenAI | frontier | 2.5 | 15 | — | — |
| gpt-5.4-mini | OpenAI | distilled | 0.75 | 4.5 | — | — |
| gpt-5.4-nano | OpenAI | distilled | 0.2 | 1.25 | — | — |
| gpt-5 | OpenAI | frontier | 1.25 | 10 | — | 2025-08 |
| gpt-5-mini | OpenAI | distilled | 0.25 | 2 | — | 2025-08 |
| gpt-5-nano | OpenAI | distilled | 0.05 | 0.4 | — | 2025-08 |
| gpt-4o | OpenAI | frontier | 2.5 | 10 | — | 2024-05 |
| gpt-4o-mini | OpenAI | distilled | 0.15 | 0.6 | — | 2024-07 |
| o1 | OpenAI | frontier | 15 | 60 | — | 2024-12 |
| o3 | OpenAI | frontier | 2 | 8 | — | 2025-04 |
| o4-mini | OpenAI | distilled | 1.1 | 4.4 | — | 2025-04 |
| Claude Fable 5.1 | Anthropic | frontier | 10 | 50 | — | — |
| Claude Opus 5 | Anthropic | frontier | 5 | 25 | — | — |
| Claude Opus 4.1 | Anthropic | frontier | 15 | 75 | — | 2025-08 |
| Claude Sonnet 5 | Anthropic | frontier | 2 | 10 | — | — |
| Claude Sonnet 4.6 | Anthropic | frontier | 3 | 15 | — | — |
| Claude Haiku 4.5 | Anthropic | distilled | 1 | 5 | — | 2025-10 |
| Claude Haiku 3.5 | Anthropic | distilled | 0.8 | 4 | — | 2024-11 |
| Gemini 3.1 Pro Preview | frontier | 2 | 12 | — | — | |
| Gemini 3.8 Flash | distilled | 0.75 | 3.75 | — | — | |
| Gemini 3.5 Flash | distilled | 1.5 | 9 | — | — | |
| Gemini 3.5 Flash-Lite | distilled | 0.3 | 2.5 | — | — | |
| Gemini 3.1 Flash-Lite | distilled | 0.25 | 1.5 | — | — | |
| Gemini 2.5 Pro | frontier | 1.25 | 10 | — | 2025-03 | |
| Gemini 2.5 Flash | distilled | 0.3 | 2.5 | — | 2025-04 | |
| Gemini 2.5 Flash-Lite | distilled | 0.1 | 0.4 | — | 2025-06 | |
| deepseek-v4-pro (peak) | DeepSeek | frontier | 1.32 | 3.96 | — | 2026-08 |
| deepseek-v4-pro (off-peak) | DeepSeek | frontier | 0.66 | 1.98 | — | 2026-08 |
| deepseek-v4-flash (peak) | DeepSeek | distilled | 0.44 | 1.32 | — | 2026-07 |
| deepseek-v4-flash (off-peak) | DeepSeek | distilled | 0.22 | 0.66 | — | 2026-07 |
| deepseek-v4-flash (pre-16 Aug 2026) | DeepSeek | distilled | 0.14 | 0.28 | — | 2026-07 |
| deepseek-v4-pro (pre-16 Aug 2026) | DeepSeek | frontier | 0.435 | 0.87 | — | 2025-10 |
| Mistral Medium 3.5 | Mistral AI | frontier | 1.5 | 7.5 | — | — |
| Mistral Large 3 | Mistral AI | frontier | 0.5 | 1.5 | — | — |
| Mistral Small 4 | Mistral AI | distilled | 0.15 | 0.6 | — | — |
| Ministral 3 (3B) | Mistral AI | distilled | 0.1 | 0.1 | 3 | — |
| Ministral 3 (8B) | Mistral AI | distilled | 0.15 | 0.15 | 8 | — |
| Ministral 3 (14B) | Mistral AI | distilled | 0.2 | 0.2 | 14 | — |
| Codestral | Mistral AI | distilled | 0.3 | 0.9 | — | — |
| grok-4.6 (<200k) | xAI | frontier | 2 | 6 | — | 2026-08 |
| grok-4.5 (<200k) | xAI | frontier | 2 | 6 | — | — |
| grok-4.3 (<200k) | xAI | frontier | 1.25 | 2.5 | — | — |
| grok-build-0.1 (<200k) | xAI | distilled | 1 | 2 | — | — |
| qwen3.8-max | Alibaba | frontier | 2 | 6 | — | — |
| qwen3.7-plus | Alibaba | frontier | 0.4 | 1.6 | — | — |
| qwen3.8-flash | Alibaba | distilled | 0.15 | 0.47 | — | — |
| qwen-turbo | Alibaba | distilled | 0.05 | 0.2 | — | — |
| qwen3.8-27b | Alibaba | open | 0.5 | 3 | 27 | — |
| qwen3-8b | Alibaba | open | 0.18 | 0.7 | 8 | 2025-04 |
| Llama 3.3 70B (Together AI) | Meta / Together AI | open | 1.04 | 1.04 | 70 | 2024-12 |
| Llama 3 8B Instruct Lite (Together AI) | Meta / Together AI | open | 0.14 | 0.14 | 8 | 2024-04 |
| Qwen2.5 7B Instruct Turbo (Together AI) | Alibaba / Together AI | open | 0.3 | 0.3 | 7 | 2024-09 |
| gpt-oss-120b (Together AI) | OpenAI / Together AI | open | 0.15 | 0.6 | 117 | 2025-08 |
| gpt-oss-120b (Groq) | OpenAI / Groq | open | 0.15 | 0.6 | 117 | 2025-08 |
| gpt-oss-20b (Groq) | OpenAI / Groq | open | 0.075 | 0.3 | 20.9 | 2025-08 |
| Qwen3.8-27B (Groq) | Alibaba / Groq | open | 0.8 | 4 | 27 | — |
| GLM-5.3-Flash (Together AI) | Zhipu / Together AI | open | 0.15 | 0.5 | — | — |
Prices are list prices in US dollars per million tokens.
Sources · 78 sources
Every figure on this page comes from one of these primary sources. Compiled 4 September 2026.
- OpenAI API pricing
- OpenAI API models reference
- GPT-4o mini: advancing cost-efficient intelligence
- OpenAI GPT-5 System Card
- OpenAI Services Agreement
- GPT-5.6 Sol model page
- GPT-5.6 Terra model page
- GPT-5.6 Luna model page
- GPT-5 mini model page
- GPT-5 nano model page
- GPT-4o model page
- o4-mini model page
- OpenAI ships GPT-5.4 mini and nano
- o4-mini: tests, features, o3 comparison, benchmarks
- gpt-oss-20b model card
- Claude platform pricing
- Claude models overview
- Claude Opus 5 model page
- Introducing Claude Sonnet 5
- Introducing Claude Haiku 4.5
- Anthropic Commercial Terms of Service
- Claude Opus 5 by Anthropic: benchmarks and pricing
- Claude benchmarks 2026
- Anthropic launches Opus 5
- Gemini API pricing
- Gemini API models
- Gemini 3.1 Pro model page
- Gemini Flash model page
- Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context and Next Generation Agentic Capabilities
- Gemma 4 model card
- Gemma 4
- gemma-3-27b-it model card
- Gemini 3.8 Flash rolling out three weeks after last release
- Gemini 3.8 Flash model stats
- Gemini 3.1 Flash-Lite benchmark results
- DeepSeek API pricing
- DeepSeek-R1 model card (with distilled model evaluations)
- DeepSeek-R1-0528 model card
- DeepSeek-R1-Distill-Qwen-32B model card
- DeepSeek-V4-Pro model card
- DeepSeek-V4-Flash model card
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- DeepSeek raises some V4 prices by more than 10x
- Llama 4 model card
- Llama 3.3 model card
- Llama 3.2 model card
- Qwen3 Technical Report
- Qwen3.8-27B model card
- Qwen3-4B-Instruct-2507 model card
- Alibaba Model Studio model pricing
- phi-4 model card
- Phi-4-mini-instruct model card
- SmolLM3-3B model card
- Mistral AI API pricing
- Mistral models overview
- Ministral-3-14B-Instruct-2512 model card
- Mistral Small 4 model page
- What is Amazon Nova? (teacher/student distillation matrix)
- What’s new in Amazon Nova 2
- The Amazon Nova Family of Models: Technical Report and Model Card
- Amazon Bedrock Model Distillation
- Amazon Bedrock pricing
- Nova Micro API pricing
- Nova Lite API pricing
- Nova 2 Lite API pricing
- Amazon Nova Pro: AWS Bedrock model guide, specs and pricing (2026)
- Together AI pricing
- Groq supported models and pricing
- xAI models and pricing
- GPT-5.6 Luna (low) vs Claude 4.5 Haiku (reasoning)
- Gemini 3.5 Flash-Lite vs GPT-5.6 Luna (high)
- Claude Sonnet 5 vs Claude Opus 5
- DeepSeek V4 Flash vs DeepSeek V4 Pro
- gpt-oss-120B vs Llama 4 Maverick
- Comparison of AI models across intelligence, performance and price
- Enterprise AI costs hit 2026 low driven by price wars and Chinese open-source models
- OpenAI discounts GPT-5.6 Luna and Terra
- The Llama 4 herd: the beginning of a new era of natively multimodal AI innovation