Updated Oct 9, 2026. 142 models ranked, 58 benchmarks, 29 sources

The superintelligence leaderboard

Every frontier AI model ranked on one score built only from public benchmarks, with every number traced to its source.

Top 10 AI models

Ranked by SI Score, 0 to 100, from 58 public benchmarks. Pillar cells glow brighter with higher scores. How the score works

All 142 ranked models
Top 10 AI models by SI Score, with pillar scores, price, context window and release date
# Model SI Score CodingMathReasoningPreference Priceper 1M, blended Context Released
1 Claude Fable 5.1 Anthropic 80.2 Coding 70 Math 91 Reasoning 87 Preference 82 $20.00 1M Sep 1, 2026
2 Claude Opus 5.5 Anthropic 78.4 Coding 72 Math 92 Reasoning 91 Preference 83 $8.00 1M Sep 22, 2026
3 GPT-6 Astra OpenAI 77.5 Coding 65 Math 93 Reasoning 87 Preference 77 $20.00 1.1M Sep 3, 2026
4 Claude Fable 5 Anthropic 76.8 Coding 70 Math 91 Reasoning 89 Preference 81 $20.00 1M Jun 9, 2026
5 Claude Opus 5 Anthropic 74.9 Coding 70 Math 87 Reasoning 84 Preference 82 $10.00 1M Jul 24, 2026
6 GPT-6.1 Sol OpenAI 73.0 Coding 65 Math 93 Reasoning 88 Preference 77 $4.00 1.1M Sep 29, 2026
7 GPT-5.6 Sol OpenAI 72.1 Coding 64 Math 89 Reasoning 82 Preference 78 $8.00 1.1M Jul 9, 2026
8 Claude Opus 4.7 Anthropic 71.2 Coding 66 Math 78 Reasoning 74 Preference 80 $10.00 1M Apr 16, 2026
9 Kimi K3 Moonshot AIOpen weights 71.0 Coding 71 Math 74 Reasoning 81 Preference 80 $6.00 1M Jul 16, 2026
10 Claude Opus 4.6 Anthropic 70.8 Coding 62 Math 74 Reasoning 79 Preference 82 $10.00 1M Feb 5, 2026

Intelligence index

SI Score of the top 25 models, colored by lab. Hatched bars have open weights.

50 70 90 #1 Claude Fable 5.1 (Anthropic): SI 80.2 80.2 Claude Fable 5.1 #2 Claude Opus 5.5 (Anthropic): SI 78.4 78.4 Claude Opus 5.5 #3 GPT-6 Astra (OpenAI): SI 77.5 77.5 GPT-6 Astra #4 Claude Fable 5 (Anthropic): SI 76.8 76.8 Claude Fable 5 #5 Claude Opus 5 (Anthropic): SI 74.9 74.9 Claude Opus 5 #6 GPT-6.1 Sol (OpenAI): SI 73.0 73.0 GPT-6.1 Sol #7 GPT-5.6 Sol (OpenAI): SI 72.1 72.1 GPT-5.6 Sol #8 Claude Opus 4.7 (Anthropic): SI 71.2 71.2 Claude Opus 4.7 #9 Kimi K3 (Moonshot AI): SI 71.0, open weights 71.0 Kimi K3 #10 Claude Opus 4.6 (Anthropic): SI 70.8 70.8 Claude Opus 4.6 #11 Muse Spark 1.3 (Meta): SI 70.1 70.1 Muse Spark 1.3 #12 Gemini 3.7 Flash (Google): SI 70.0 70.0 Gemini 3.7 Flash #13 GPT-5.4 (OpenAI): SI 69.7 69.7 GPT-5.4 #14 GPT-5.5 (OpenAI): SI 69.5 69.5 GPT-5.5 #15 Claude Sonnet 5.5 (Anthropic): SI 69.4 69.4 Claude Sonnet 5.5 #16 GPT-6 Sol (OpenAI): SI 68.9 68.9 GPT-6 Sol #17 Gemini 3.8 Flash (Google): SI 68.8 68.8 Gemini 3.8 Flash #18 Qwen3.8 Max (Alibaba / Qwen): SI 68.7 68.7 Qwen3.8 Max #19 Gemini 3.5 Flash (Google): SI 68.3 68.3 Gemini 3.5 Flash #20 Claude Sonnet 4.6 (Anthropic): SI 68.3 68.3 Claude Sonnet 4.6 #21 GPT-5.6 Terra (OpenAI): SI 68.2 68.2 GPT-5.6 Terra #22 Gemini 3.1 Pro Preview (Google): SI 68.1 68.1 Gemini 3.1 Pro Preview #23 MiniMax-M3 (MiniMax): SI 68.0, open weights 68.0 MiniMax-M3 #24 GLM-5.3 (Z.ai): SI 67.7, open weights 67.7 GLM-5.3 #25 Claude Sonnet 5 (Anthropic): SI 67.7 67.7 Claude Sonnet 5 50 70 90 #1 Claude Fable 5.1 (Anthropic): SI 80.2 80.2 Claude Fable 5.1 #2 Claude Opus 5.5 (Anthropic): SI 78.4 78.4 Claude Opus 5.5 #3 GPT-6 Astra (OpenAI): SI 77.5 77.5 GPT-6 Astra #4 Claude Fable 5 (Anthropic): SI 76.8 76.8 Claude Fable 5 #5 Claude Opus 5 (Anthropic): SI 74.9 74.9 Claude Opus 5 #6 GPT-6.1 Sol (OpenAI): SI 73.0 73.0 GPT-6.1 Sol #7 GPT-5.6 Sol (OpenAI): SI 72.1 72.1 GPT-5.6 Sol #8 Claude Opus 4.7 (Anthropic): SI 71.2 71.2 Claude Opus 4.7 #9 Kimi K3 (Moonshot AI): SI 71.0, open weights 71.0 Kimi K3 #10 Claude Opus 4.6 (Anthropic): SI 70.8 70.8 Claude Opus 4.6 #11 Muse Spark 1.3 (Meta): SI 70.1 70.1 Muse Spark 1.3 #12 Gemini 3.7 Flash (Google): SI 70.0 70.0 Gemini 3.7 Flash
AnthropicOpenAIMoonshot AIMetaGoogleAlibaba / QwenMiniMaxZ.ai Open weights
Open the full leaderboard

Best by task

Pillar leaders, among models with at least three results. Price is blended per 1M tokens.

  1. 1DeepSeek V4.1 FlashDeepSeek 73.3$0.26
  2. 2Claude Opus 5.5Anthropic 72.5$8.00
  3. 3Kimi K3Moonshot AI 71.1$6.00
  4. 4Claude Fable 5Anthropic 70.2$20.00
  5. 5Claude Fable 5.1Anthropic 69.8$20.00
  6. 6Claude Opus 5Anthropic 69.7$10.00
  7. 7Hy3Tencent 69.2$0.00
  8. 8Qwen3.8 MaxAlibaba / Qwen 69.1$2.48
Full coding ranking
  1. 1GPT-6.1 SolOpenAI 93.0$4.00
  2. 2GPT-6 AstraOpenAI 92.9$20.00
  3. 3Claude Opus 5.5Anthropic 91.5$8.00
  4. 4Claude Fable 5.1Anthropic 91.5$20.00
  5. 5GPT-6 SolOpenAI 91.3$4.00
  6. 6Claude Fable 5Anthropic 91.0$20.00
  7. 7Claude Sonnet 5.5Anthropic 90.5$4.00
  8. 8GPT-5.6 SolOpenAI 89.3$8.00
Full math ranking
  1. 1Claude Opus 5.5Anthropic 91.4$8.00
  2. 2Muse Spark 1.3Meta 90.5$2.00
  3. 3Muse Spark 1.2Meta 90.4$2.00
  4. 4Claude Fable 5Anthropic 89.4$20.00
  5. 5Claude Sonnet 5.5Anthropic 88.6$4.00
  6. 6GPT-6.1 SolOpenAI 87.5$4.00
  7. 7Claude Sonnet 5Anthropic 87.3$4.00
  8. 8Claude Fable 5.1Anthropic 87.1$20.00
Full reasoning ranking

Intelligence vs price

Blended API price per 1M tokens, log scale. Up and left is more score per dollar.

20 40 60 80 100 $0.1$0.5$2.5$10$50 USD / 1M tokens (log scale; 3:1 in/out) Claude Fable 5.1 — SI 80.2 · $20.00/1M blended · confidence 100% Claude Opus 5.5 — SI 78.4 · $8.00/1M blended · confidence 100% GPT-6 Astra — SI 77.5 · $20.00/1M blended · confidence 100% Claude Fable 5 — SI 76.8 · $20.00/1M blended · confidence 100% Claude Opus 5 — SI 74.9 · $10.00/1M blended · confidence 100% GPT-6.1 Sol — SI 73.0 · $4.00/1M blended · confidence 100% GPT-5.6 Sol — SI 72.1 · $8.00/1M blended · confidence 100% Claude Opus 4.7 — SI 71.2 · $10.00/1M blended · confidence 100% Kimi K3 — SI 71.0 · $6.00/1M blended · confidence 100% Claude Opus 4.6 — SI 70.8 · $10.00/1M blended · confidence 100% Muse Spark 1.3 — SI 70.1 · $2.00/1M blended · confidence 93% Gemini 3.7 Flash — SI 70.0 · $1.50/1M blended · confidence 100% GPT-5.4 — SI 69.7 · $5.63/1M blended · confidence 100% GPT-5.5 — SI 69.5 · $11.25/1M blended · confidence 100% Claude Sonnet 5.5 — SI 69.4 · $4.00/1M blended · confidence 93% GPT-6 Sol — SI 68.9 · $4.00/1M blended · confidence 100% Gemini 3.8 Flash — SI 68.8 · $1.50/1M blended · confidence 100% Qwen3.8 Max — SI 68.7 · $2.48/1M blended · confidence 85% Gemini 3.5 Flash — SI 68.3 · $3.38/1M blended · confidence 100% Claude Sonnet 4.6 — SI 68.3 · $6.00/1M blended · confidence 100% GPT-5.6 Terra — SI 68.2 · $4.50/1M blended · confidence 100% Gemini 3.1 Pro Preview — SI 68.1 · $4.50/1M blended · confidence 100% MiniMax-M3 — SI 68.0 · $0.52/1M blended · confidence 100% GLM-5.3 — SI 67.7 · $2.15/1M blended · confidence 93% Claude Sonnet 5 — SI 67.7 · $4.00/1M blended · confidence 93% Claude Opus 4.8 — SI 67.5 · $10.00/1M blended · confidence 100% Grok 4.6 — SI 66.6 · $3.00/1M blended · confidence 100% Qwen3.6 Max Preview — SI 66.4 · $2.92/1M blended · confidence 64% GLM-5.3-Flash — SI 66.2 · $0.24/1M blended · confidence 100% DeepSeek V4.1 Flash — SI 66.2 · $0.26/1M blended · confidence 69% GPT-5.2 — SI 65.9 · $4.81/1M blended · confidence 100% GLM-5.2 — SI 65.7 · $2.15/1M blended · confidence 100% Gemini 3.6 Flash — SI 65.5 · $1.50/1M blended · confidence 100% Grok 4.7 — SI 65.5 · $3.00/1M blended · confidence 100% Grok 4.5 — SI 64.8 · $3.00/1M blended · confidence 100% Kimi K2.6 — SI 64.0 · $1.71/1M blended · confidence 85% Qwen3.5 397B-A17B — SI 64.0 · $0.39/1M blended · confidence 69% GPT-5.5 Pro — SI 63.5 · $67.50/1M blended · confidence 53% GPT-5.6 Luna — SI 63.4 · $0.45/1M blended · confidence 100% GLM-5.1 — SI 63.3 · $2.15/1M blended · confidence 64% DeepSeek V4 Pro — SI 63.1 · $1.98/1M blended · confidence 85% DeepSeek V4 Pro 0813 — SI 63.0 · $0.99/1M blended · confidence 69% Muse Spark 1.1 — SI 62.9 · $2.00/1M blended · confidence 61% Qwen3.7 Plus — SI 62.7 · $0.48/1M blended · confidence 64% Qwen3.5 35B-A3B — SI 62.7 · $0.16/1M blended · confidence 64% Claude Opus 4.5 — SI 62.6 · $10.00/1M blended · confidence 100% MiMo-V2.5-Pro — SI 62.6 · $0.54/1M blended · confidence 88% Muse Spark 1.2 — SI 62.3 · $2.00/1M blended · confidence 53% Mistral Large 4 — SI 62.2 · $1.03/1M blended · confidence 64% Qwen3.8 27B — SI 61.9 · $1.13/1M blended · confidence 69% GPT-6 Luna — SI 61.6 · $0.20/1M blended · confidence 100% Qwen3.6 Plus — SI 61.6 · $0.62/1M blended · confidence 80% GLM-4.7 — SI 61.3 · $1.00/1M blended · confidence 69% GPT-5.4 Pro — SI 61.3 · $67.50/1M blended · confidence 69% Claude Haiku 4.5 — SI 60.5 · $2.00/1M blended · confidence 64% Qwen3.5 Flash — SI 59.5 · $0.09/1M blended · confidence 64% Claude Haiku 5.5 — SI 59.4 · $0.20/1M blended · confidence 68% GLM-5 — SI 59.4 · $1.55/1M blended · confidence 90% Kimi K2.5 — SI 59.0 · $1.20/1M blended · confidence 100% Qwen3.7 Max — SI 58.6 · $2.48/1M blended · confidence 53% MiniMax-M2.7 — SI 57.9 · $0.52/1M blended · confidence 88% MiniMax-M2.5 — SI 57.8 · $0.52/1M blended · confidence 100% Grok 4.20 (Reasoning) — SI 57.7 · $1.56/1M blended · confidence 80% MiMo-V2.6-Pro — SI 57.4 · $0.54/1M blended · confidence 88% GPT-5.4 nano — SI 57.4 · $0.46/1M blended · confidence 100% GPT-5.4 mini — SI 56.9 · $1.69/1M blended · confidence 100% MiMo-V2.5 — SI 56.4 · $0.18/1M blended · confidence 88% DeepSeek V4 Flash — SI 56.2 · $0.52/1M blended · confidence 53% MiMo-V2.6-Flash — SI 55.8 · $0.18/1M blended · confidence 88% GPT-5 — SI 55.7 · $3.44/1M blended · confidence 100% MiniMax-M2 — SI 55.7 · $0.52/1M blended · confidence 96% GPT-5.1 — SI 55.7 · $3.44/1M blended · confidence 85% Qwen3.6 27B — SI 55.6 · $1.35/1M blended · confidence 53% Claude Opus 4 — SI 55.5 · $30.00/1M blended · confidence 100% o3 — SI 55.4 · $3.50/1M blended · confidence 87% Gemini 3.5 Flash Lite — SI 55.2 · $0.85/1M blended · confidence 100% Gemini 3 Flash Preview — SI 54.8 · $1.13/1M blended · confidence 93% Grok 4.3 — SI 54.6 · $1.56/1M blended · confidence 80% DeepSeek-V3 — SI 54.1 · $0.48/1M blended · confidence 93% Claude Sonnet 4.5 — SI 53.9 · $6.00/1M blended · confidence 80% DeepSeek V3.2 — SI 53.3 · $0.31/1M blended · confidence 56% Step 3.5 Flash — SI 53.0 · $0.15/1M blended · confidence 88% Qwen3 30B A3B — SI 52.6 · $0.19/1M blended · confidence 64% o4-mini — SI 51.2 · $1.93/1M blended · confidence 87% GPT-5 Mini — SI 50.6 · $0.69/1M blended · confidence 99% Qwen3 32B — SI 50.4 · $0.28/1M blended · confidence 67% Qwen3 235B-A22B — SI 50.3 · $0.50/1M blended · confidence 76% Qwen3 Max — SI 50.2 · $0.63/1M blended · confidence 69% GPT-4o (2024-05-13) — SI 49.4 · $7.50/1M blended · confidence 93% DeepSeek-R1 — SI 47.9 · $0.96/1M blended · confidence 88% GLM-4.6 — SI 47.8 · $1.00/1M blended · confidence 55% Claude Sonnet 4 — SI 46.6 · $6.00/1M blended · confidence 100% GPT-5 Pro — SI 45.8 · $41.25/1M blended · confidence 64% Gemini 2.5 Pro — SI 45.2 · $3.44/1M blended · confidence 90% GPT-4o (2024-11-20) — SI 42.2 · $4.38/1M blended · confidence 61% o3-mini — SI 41.6 · $1.93/1M blended · confidence 96% GPT-4.1 — SI 41.5 · $3.50/1M blended · confidence 100% Claude Haiku 3.5 — SI 40.8 · $1.60/1M blended · confidence 100% GPT-5 Nano — SI 39.1 · $0.14/1M blended · confidence 85% GPT-4.1 mini — SI 36.9 · $0.70/1M blended · confidence 87% GPT-4o (2024-08-06) — SI 36.7 · $4.38/1M blended · confidence 100% GPT-4.1 nano — SI 32.1 · $0.17/1M blended · confidence 82% GPT-4o — SI 31.2 · $4.38/1M blended · confidence 59% GPT-4o mini — SI 22.9 · $0.26/1M blended · confidence 100%
Best score at its price or lower Other ranked models (fainter means lower confidence) Open weights
Open pricing

Labs

Each lab's best ranked model.

  1. AnthropicClaude Fable 5.1 80.221 ranked
  2. OpenAIGPT-6 Astra 77.533 ranked
  3. Moonshot AIKimi K3 71.04 ranked
  4. MetaMuse Spark 1.3 70.19 ranked
  5. GoogleGemini 3.7 Flash 70.014 ranked
  6. Alibaba / QwenQwen3.8 Max 68.716 ranked
  7. MiniMaxMiniMax-M3 68.04 ranked
  8. Z.aiGLM-5.3 67.78 ranked
All labs

Latest releases

Newest text models in the catalog.

All releases

Price, top 10

Blended price per 1M tokens (3 input : 1 output).

  1. Claude Fable 5.1Anthropic $20.0080.2
  2. Claude Opus 5.5Anthropic $8.0078.4
  3. GPT-6 AstraOpenAI $20.0077.5
  4. Claude Fable 5Anthropic $20.0076.8
  5. Claude Opus 5Anthropic $10.0074.9
  6. GPT-6.1 SolOpenAI $4.0073.0
  7. GPT-5.6 SolOpenAI $8.0072.1
  8. Claude Opus 4.7Anthropic $10.0071.2
Compare all prices

Best value

SI Score per blended dollar.

  1. Gemma 4 26B A4B ITGoogle 253.6$0.00
  2. Gemma 4 31B ITGoogle 252.0$0.00
  3. Hy3Tencent 242.8$0.00
  4. GLM-4.7-FlashZ.ai 213.2$0.00
  5. Qwen3.5 FlashAlibaba / Qwen 173.2$0.09
  6. Qwen3.5 35B-A3BAlibaba / Qwen 153.9$0.16
  7. GPT-6 LunaOpenAI 136.9$0.20
  8. GLM-5.3-FlashZ.ai 135.8$0.24
Price and performance ranking

Context window

Largest advertised input, log scale.

  1. Llama 4 Scout 17B InstructMeta 10M51.0
  2. GPT-6 AstraOpenAI 1.1M77.5
  3. GPT-6.1 SolOpenAI 1.1M73.0
  4. GPT-5.6 SolOpenAI 1.1M72.1
  5. GPT-5.4OpenAI 1.1M69.7
  6. GPT-5.5OpenAI 1.1M69.5
  7. GPT-6 SolOpenAI 1.1M68.9
  8. GPT-5.6 TerraOpenAI 1.1M68.2
Long-context ranking

The frontier

Best SI Score among models released by each date.

20 40 60 80 20252026 Claude Haiku 3: SI 39.9, released Mar 2024GPT-4o (2024-05-13): SI 49.4, released May 2024GPT-4o: SI 31.2, released May 2024GPT-4o mini: SI 22.9, released Jul 2024Llama-3.1-70B-Instruct: SI 45.7, released Jul 2024Llama-3.1-8B-Instruct: SI 37.7, released Jul 2024GPT-4o (2024-08-06): SI 36.7, released Aug 2024Llama-3.2-1B: SI 32.1, released Sep 2024Claude Sonnet 3.5 v2: SI 51.9, released Oct 2024Claude Haiku 3.5: SI 40.8, released Oct 2024Qwen2.5-Coder-32B-Instruct: SI 39.6, released Nov 2024Mistral Large 2.1: SI 35.3, released Nov 2024GPT-4o (2024-11-20): SI 42.2, released Nov 2024Nova Pro: SI 37.4, released Dec 2024Nova Lite: SI 36.6, released Dec 2024Llama-3.3-70B-Instruct: SI 32.7, released Dec 2024o3-mini: SI 41.6, released Dec 2024DeepSeek-V3: SI 54.1, released Dec 2024DeepSeek-R1: SI 47.9, released Jan 2025Claude Sonnet 3.7: SI 48.9, released Feb 2025QwQ 32B: SI 48.2, released Mar 2025Gemma 3 12B IT: SI 45.1, released Mar 2025Gemma 3 27B IT: SI 42.0, released Mar 2025Gemma 3 4B IT: SI 39.2, released Mar 2025Mistral Small 3.1 24B: SI 48.4, released Mar 2025DeepSeek V3 0324: SI 55.7, released Mar 2025Qwen3 32B: SI 50.4, released Apr 2025Qwen3 235B-A22B: SI 50.3, released Apr 2025Llama 4 Scout 17B Instruct: SI 51.0, released Apr 2025Llama 4 Maverick 17B Instruct: SI 30.0, released Apr 2025GPT-4.1: SI 41.5, released Apr 2025GPT-4.1 mini: SI 36.9, released Apr 2025GPT-4.1 nano: SI 32.1, released Apr 2025o3: SI 55.4, released Apr 2025o4-mini: SI 51.2, released Apr 2025Qwen3 30B A3B: SI 52.6, released Apr 2025Mistral Medium 3: SI 33.0, released May 2025Claude Opus 4: SI 55.5, released May 2025Claude Sonnet 4: SI 46.6, released May 2025Gemini 2.5 Pro: SI 45.2, released Jun 2025GPT OSS 120B: SI 54.8, released Aug 2025GPT OSS 20B: SI 52.5, released Aug 2025Claude Opus 4.1: SI 51.5, released Aug 2025GPT-5: SI 55.7, released Aug 2025GPT-5 Mini: SI 50.6, released Aug 2025GPT-5 Nano: SI 39.1, released Aug 2025Qwen3 Max: SI 50.2, released Sep 2025Claude Sonnet 4.5: SI 53.9, released Sep 2025GLM-4.6: SI 47.8, released Sep 2025GPT-5 Pro: SI 45.8, released Oct 2025Claude Haiku 4.5: SI 60.5, released Oct 2025MiniMax-M2: SI 55.7, released Oct 2025Claude Opus 4.5: SI 62.6, released Nov 2025Kimi K2 Thinking Turbo: SI 66.1, released Nov 2025GPT-5.1: SI 55.7, released Nov 2025Gemini 3 Pro Preview: SI 58.6, released Nov 2025DeepSeek V3.2: SI 53.3, released Dec 2025GPT-5.2: SI 65.9, released Dec 2025Gemini 3 Flash Preview: SI 54.8, released Dec 2025GLM-4.7: SI 61.3, released Dec 2025Kimi K2.5: SI 59.0, released Jan 2026GLM-4.7-Flash: SI 53.3, released Jan 2026Step 3.5 Flash: SI 53.0, released Jan 2026Claude Opus 4.6: SI 70.8, released Feb 2026GLM-5: SI 59.4, released Feb 2026MiniMax-M2.5: SI 57.8, released Feb 2026Qwen3.5 397B-A17B: SI 64.0, released Feb 2026Claude Sonnet 4.6: SI 68.3, released Feb 2026Gemini 3.1 Pro Preview: SI 68.1, released Feb 2026Qwen3.5 35B-A3B: SI 62.7, released Feb 2026Qwen3.5 Flash: SI 59.5, released Feb 2026GPT-5.4: SI 69.7, released Mar 2026GPT-5.4 Pro: SI 61.3, released Mar 2026Grok 4.20 (Reasoning): SI 57.7, released Mar 2026GPT-5.4 nano: SI 57.4, released Mar 2026GPT-5.4 mini: SI 56.9, released Mar 2026MiniMax-M2.7: SI 57.9, released Mar 2026Gemma 4 26B A4B IT: SI 63.4, released Apr 2026Gemma 4 31B IT: SI 63.0, released Apr 2026Qwen3.6 Plus: SI 61.6, released Apr 2026GLM-5.1: SI 63.3, released Apr 2026Muse Spark 1.1: SI 62.9, released Apr 2026Claude Opus 4.7: SI 71.2, released Apr 2026Grok 4.3: SI 54.6, released Apr 2026Qwen3.6 Max Preview: SI 66.4, released Apr 2026Kimi K2.6: SI 64.0, released Apr 2026MiMo-V2.5-Pro: SI 62.6, released Apr 2026MiMo-V2.5: SI 56.4, released Apr 2026Qwen3.6 27B: SI 55.6, released Apr 2026GPT-5.5: SI 69.5, released Apr 2026GPT-5.5 Pro: SI 63.5, released Apr 2026DeepSeek V4 Pro: SI 63.1, released Apr 2026DeepSeek V4 Flash: SI 56.2, released Apr 2026GPT-5.5 Instant: SI 58.3, released May 2026Gemini 3.5 Flash: SI 68.3, released May 2026Qwen3.7 Max: SI 58.6, released May 2026Claude Opus 4.8: SI 67.5, released May 2026MiniMax-M3: SI 68.0, released Jun 2026Qwen3.7 Plus: SI 62.7, released Jun 2026Nemotron 3 Ultra 550B A55B: SI 61.2, released Jun 2026Claude Fable 5: SI 76.8, released Jun 2026GLM-5.2: SI 65.7, released Jun 2026Fugu Ultra: SI 57.9, released Jun 2026Fugu: SI 55.9, released Jun 2026Claude Sonnet 5: SI 67.7, released Jun 2026LongCat-2.0: SI 56.6, released Jun 2026Hy3: SI 60.7, released Jul 2026Grok 4.5: SI 64.8, released Jul 2026GPT-5.6 Sol: SI 72.1, released Jul 2026GPT-5.6 Terra: SI 68.2, released Jul 2026GPT-5.6 Luna: SI 63.4, released Jul 2026Inkling: SI 66.5, released Jul 2026Kimi K3: SI 71.0, released Jul 2026Gemini 3.6 Flash: SI 65.5, released Jul 2026Gemini 3.5 Flash Lite: SI 55.2, released Jul 2026Claude Opus 5: SI 74.9, released Jul 2026Inkling Small: SI 53.6, released Jul 2026DeepSeek V4 Flash 0731: SI 61.0, released Jul 2026Qwen3.8 Max: SI 68.7, released Aug 2026Muse Spark 1.2: SI 62.3, released Aug 2026Nemotron 3.5 Lightning 30B A3B: SI 46.1, released Aug 2026Grok 4.6: SI 66.6, released Aug 2026DeepSeek V4 Pro 0813: SI 63.0, released Aug 2026Gemini 3.7 Flash: SI 70.0, released Aug 2026GLM-5.3: SI 67.7, released Aug 2026Qwen3.8 27B: SI 61.9, released Aug 2026GLM-5.3-Flash: SI 66.2, released Aug 2026Claude Fable 5.1: SI 80.2, released Sep 2026Muse Spark 1.3: SI 70.1, released Sep 2026Gemini 3.8 Flash: SI 68.8, released Sep 2026GPT-6 Astra: SI 77.5, released Sep 2026DeepSeek V4.1 Flash: SI 66.2, released Sep 2026Grok 4.7: SI 65.5, released Sep 2026Claude Opus 5.5: SI 78.4, released Sep 2026GPT-6 Sol: SI 68.9, released Sep 2026GPT-6 Luna: SI 61.6, released Sep 2026MiMo-V2.6-Pro: SI 57.4, released Sep 2026MiMo-V2.6-Flash: SI 55.8, released Sep 2026Claude Sonnet 5.5: SI 69.4, released Sep 2026GPT-6.1 Sol: SI 73.0, released Sep 2026Mistral Large 4: SI 62.2, released Oct 2026Claude Haiku 5.5: SI 59.4, released Oct 2026 Claude Opus 4.6DeepSeek V3 0324GPT-4o (2024-05-13) Claude Fable 5.1 SI 80.2, Sep 2026 20 40 60 80 20252026 Claude Fable 5.1 80.2
Release timeline

AI news

Model launches from the labs and moves on our leaderboard.

All AI news

Open weights vs proprietary

The best downloadable model trails the best closed model by 9.2 points.

40 60 80 20252026 Claude Haiku 3: 39.9GPT-4o (2024-05-13): 49.4Claude Sonnet 3.5 v2: 51.9o3: 55.4Claude Opus 4: 55.5GPT-5: 55.7Claude Haiku 4.5: 60.5Claude Opus 4.5: 62.6GPT-5.2: 65.9Claude Opus 4.6: 70.8Claude Opus 4.7: 71.2Claude Fable 5: 76.8Claude Fable 5.1: 80.2 Llama-3.1-70B-Instruct: 45.7DeepSeek-V3: 54.1DeepSeek V3 0324: 55.7Kimi K2 Thinking Turbo: 66.1MiniMax-M3: 68.0Kimi K3: 71.0 80.2Claude Fable 5.1 71.0Kimi K3 40 60 80 20252026 Claude Haiku 3: 39.9GPT-4o (2024-05-13): 49.4Claude Sonnet 3.5 v2: 51.9o3: 55.4Claude Opus 4: 55.5GPT-5: 55.7Claude Haiku 4.5: 60.5Claude Opus 4.5: 62.6GPT-5.2: 65.9Claude Opus 4.6: 70.8Claude Opus 4.7: 71.2Claude Fable 5: 76.8Claude Fable 5.1: 80.2 Llama-3.1-70B-Instruct: 45.7DeepSeek-V3: 54.1DeepSeek V3 0324: 55.7Kimi K2 Thinking Turbo: 66.1MiniMax-M3: 68.0Kimi K3: 71.0 80.2Claude Fable… 71.0Kimi K3
Best proprietary Best open weights
Best open-weight models

Benchmarks

Most-reported benchmarks and who leads each.

All benchmarks

Hear when the frontier moves

A notification on this device when a new model takes #1 or enters the top 10. No account, no email; turn it off any time.

What to be alerted about
RSS feed

Understand the numbers

Methodology

How the SI Score is built, what the confidence % means, and which sources feed it, with their licenses.

What is superintelligence?

The research term, the corporate lane name and the new government spelling, untangled and dated.

“Super Intelligence” in U.S. policy

Executive Order 14434 made it executive-branch vocabulary on Sep 29, 2026. What it says, verbatim.

Frontier models API

The current frontier per lab as free JSON, for agents and developers. No key, no signup.

What changed

All releases · RSS feed

Alerts on this device

What to be alerted about
RSS feed