GPT-6 Astra jumps to #1, Claude Opus 4.6 added (9 Sept 2026)

BenchLeader refreshed 1211 models from 19 of 19 data feeds (benchmark publishers, aggregators and price lists). 261 new models, 10 top-10 rank changes, 122 price changes, 4579 new or updated benchmark results across 19 data feeds.

New models

Claude Opus 4.6 (thinking) (Anthropic) appeared on 9 benchmarks, entering the BenchLeader Index at #59. (details)
Qwen3 6 (max) (Alibaba) appeared on 5 benchmarks, entering the BenchLeader Index at #60. (details)
Gemini 3 Flash (thinking) (Google) appeared on 6 benchmarks, entering the BenchLeader Index at #66. (details)
Grok 4.3 (medium) (xAI) appeared on 6 benchmarks, entering the BenchLeader Index at #81. (details)
Claude Opus 4.7 (no reasoning) (Anthropic) appeared on 6 benchmarks, entering the BenchLeader Index at #94. (details)
Grok 4.3 (low) (xAI) appeared on 6 benchmarks, entering the BenchLeader Index at #102. (details)
KAT-Coder-Pro V2 (KwaiKAT) appeared on 5 benchmarks, entering the BenchLeader Index at #131. (details)
Qwen3.5 Plus (thinking) (Alibaba) appeared on 6 benchmarks, entering the BenchLeader Index at #147. (details)
Gemini 3 Pro (low) (Google) appeared on 5 benchmarks, entering the BenchLeader Index at #161. (details)
Claude Sonnet 4.6 (low) (Anthropic) appeared on 6 benchmarks, entering the BenchLeader Index at #174. (details)
Claude Opus 4.6 (no reasoning) (Anthropic) appeared on 6 benchmarks, entering the BenchLeader Index at #179. (details)
Kimi K2.6 (no reasoning) (Moonshot AI) appeared on 5 benchmarks, entering the BenchLeader Index at #180. (details)
GPT-5.1 (thinking) (OpenAI) appeared on 6 benchmarks, entering the BenchLeader Index at #181. (details)
GLM-5 (no reasoning) (Zhipu AI) appeared on 5 benchmarks, entering the BenchLeader Index at #197. (details)
Command A+ (Cohere) appeared on 6 benchmarks, entering the BenchLeader Index at #213. (details)
Grok 3 Mini Fast (high) (xAI) appeared on 8 benchmarks, entering the BenchLeader Index at #222. (details)
Qwen3.5 Omni Plus (Alibaba) appeared on 6 benchmarks, entering the BenchLeader Index at #229. (details)
Kimi K2.5 (no reasoning) (Moonshot AI) appeared on 6 benchmarks, entering the BenchLeader Index at #241. (details)
Gemini 2.5 Flash 09 (thinking) (Google) appeared on 14 benchmarks, entering the BenchLeader Index at #246.
Gemma 4 12B (Google) appeared on 6 benchmarks, entering the BenchLeader Index at #258. (details)
JT-35B-Flash (China Mobile) appeared on 5 benchmarks, entering the BenchLeader Index at #259. (details)
Qwen3.5 27B (no reasoning) (Alibaba) appeared on 6 benchmarks, entering the BenchLeader Index at #262. (details)
Grok 3 mini (thinking) (xAI) appeared on 5 benchmarks, entering the BenchLeader Index at #274. (details)
Nemotron Cascade 2 30B A3B (NVIDIA) appeared on 5 benchmarks, entering the BenchLeader Index at #275. (details)
K-EXAONE (LG AI Research) appeared on 5 benchmarks, entering the BenchLeader Index at #276. (details)
Nova 2.0 Lite (high) (thinking) (Amazon) appeared on 6 benchmarks, entering the BenchLeader Index at #282. (details)
MiMo-V2.5-Pro (no reasoning) (Xiaomi) appeared on 5 benchmarks, entering the BenchLeader Index at #284. (details)
Gemma-4-31B-IT (no reasoning) (NVIDIA) appeared on 6 benchmarks, entering the BenchLeader Index at #285. (details)
GLM-4.7 (no reasoning) (Zhipu AI) appeared on 5 benchmarks, entering the BenchLeader Index at #286. (details)
Grok 3 Mini Fast (low) (xAI) appeared on 8 benchmarks, entering the BenchLeader Index at #288. (details)
…and 231 more.

Movement in the top 10

GPT-6 Astra (max) moved from #2 to #1 in the BenchLeader Index. (details)
Claude Fable 5 moved from #6 to #2 in the BenchLeader Index. (details)
Claude Fable 5.1 (max) moved from #1 to #4 in the BenchLeader Index. (details)
Claude Fable 5.1 (high) moved from #10 to #8 in the BenchLeader Index. (details)
GPT-6 Astra (high) moved from #7 to #9 in the BenchLeader Index. (details)
Claude Opus 5 (high) moved from #8 to #10 in the BenchLeader Index. (details)
Claude Opus 5 (max) moved from #4 to #16 in the BenchLeader Index. (details)
Claude Fable 5 (max) moved from #3 to #23 in the BenchLeader Index. (details)
Claude Opus 4.7 moved from #9 to #25 in the BenchLeader Index. (details)
GPT-5.6 Sol (max) moved from #5 to #26 in the BenchLeader Index. (details)

Price changes

GPT-6 Astra (max) input price moved from $10.00/M to $11.00/M (10% up). (details)
GPT-6 Astra (max) output price moved from $50.00/M to $55.00/M (10% up). (details)
GPT-5.6 Sol (high) input price moved from $4.40/M to $4.00/M (9% down). (details)
GPT-5.6 Sol (high) output price moved from $22.00/M to $20.00/M (9% down). (details)
GPT-5.6 Sol (xhigh) input price moved from $4.40/M to $4.00/M (9% down). (details)
GPT-5.6 Sol (xhigh) output price moved from $22.00/M to $20.00/M (9% down). (details)
GPT-5.5 (high) input price moved from $5.50/M to $5.00/M (9% down). (details)
GPT-5.5 (high) output price moved from $33.00/M to $30.00/M (9% down). (details)
GPT-5.6 Sol (medium) input price moved from $4.40/M to $4.00/M (9% down). (details)
GPT-5.6 Sol (medium) output price moved from $22.00/M to $20.00/M (9% down). (details)
GPT-5.5 (medium) input price moved from $5.50/M to $5.00/M (9% down). (details)
GPT-5.5 (medium) output price moved from $33.00/M to $30.00/M (9% down). (details)
GPT-5.6 Sol (low) input price moved from $4.40/M to $4.00/M (9% down). (details)
GPT-5.6 Sol (low) output price moved from $22.00/M to $20.00/M (9% down). (details)
GPT-6 Astra (no reasoning) input price moved from $10.00/M to $11.00/M (10% up). (details)
GPT-6 Astra (no reasoning) output price moved from $50.00/M to $55.00/M (10% up). (details)
GPT-5.6 Terra (xhigh) input price moved from $2.20/M to $2.00/M (9% down). (details)
GPT-5.6 Terra (xhigh) output price moved from $13.20/M to $12.00/M (9% down). (details)
Grok 4.6 (medium) input price moved from $2.20/M to $2.00/M (9% down). (details)
Grok 4.6 (medium) output price moved from $6.60/M to $6.00/M (9% down). (details)
GPT-5.6 Terra (high) input price moved from $2.20/M to $2.00/M (9% down). (details)
GPT-5.6 Terra (high) output price moved from $13.20/M to $12.00/M (9% down). (details)
Grok 4.6 (xhigh) input price moved from $2.20/M to $2.00/M (9% down). (details)
Grok 4.6 (xhigh) output price moved from $6.60/M to $6.00/M (9% down). (details)
GPT-5.5 (low) input price moved from $5.50/M to $5.00/M (9% down). (details)
GPT-5.5 (low) output price moved from $33.00/M to $30.00/M (9% down). (details)
GPT-5.4 (low) input price moved from $2.75/M to $2.50/M (9% down). (details)
GPT-5.4 (low) output price moved from $16.50/M to $15.00/M (9% down). (details)
Grok 4.6 (low) input price moved from $2.20/M to $2.00/M (9% down). (details)
Grok 4.6 (low) output price moved from $6.60/M to $6.00/M (9% down). (details)
…and 92 more.

New benchmark results

AA Intelligence Index v4.3.2: 374 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
GPQA Diamond (AA): 362 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
Humanity's Last Exam (AA): 360 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
AA-LCR: 313 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
AA-Omniscience: 311 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
IFBench: 260 new results, including Claude Fable 5, GPT-5.6 Sol, GPT-5.6 Sol… (board)
τ²-Bench Telecom (AA): 254 new results, including Claude Fable 5, GPT-5.6 Sol, GPT-5.6 Sol… (board)
Terminal-Bench Hard: 248 new results, including Claude Fable 5, GPT-5.6 Sol, GPT-5.6 Sol… (board)
MMMU-Pro: 183 new results, including Claude Opus 5, GPT-6 Astra, GPT-6 Astra… (board)
SciCode (AA): 131 new results, including Claude Fable 5, Claude Fable 5.1, Claude Opus 5… (board)
LiveCodeBench: 116 new results, including Claude Fable 5, Claude Opus 5, Claude Fable 5.1… (board)
TaxEval: 115 new results, including Claude Fable 5, Claude Opus 5, Claude Fable 5.1… (board)

Score revisions

Claude Fable 5.1 (max) on LMArena Agent: 15.9 → 14.5. (details)
Claude Opus 5 (high) on LMArena Agent: 12.7 → 11.4. (details)
GPT-5.6 Sol (xhigh) on LMArena Agent: 9.3 → 7.7. (details)
Claude Opus 5 (max) on LMArena Agent: 11.7 → 10.8. (details)
GPT-5.5 on LMArena Agent: 4.4 → 3. (details)
Muse Spark 1.1 on LMArena Agent: -1.4 → -2.6. (details)
GPT-5.6 Terra (xhigh) on LMArena Agent: 3 → 1.5. (details)
Gemini 3.1 Pro on LMArena Agent: -3.7 → -5.3. (details)
Qwen3 8 (max) on LMArena Agent: 5.5 → 3.9. (details)
Claude Fable 5 (high) on LMArena Agent: 10.2 → 9.2. (details)
GPT-5.5 (xhigh) on LMArena Agent: 7.2 → 5.3. (details)
Grok 4.5 on LMArena Agent: 5.7 → 3.9. (details)
Kimi K3 (max) on LMArena Agent: 8 → 6.6. (details)
Grok 4.6 (xhigh) on LMArena Agent: 5 → 3.2. (details)
Muse Spark 1.2 (xhigh) on LMArena Agent: 0.5 → -1.6. (details)
Claude Opus 4.8 (high) on LMArena Agent: 9.2 → 7.7. (details)
GLM-5.3-Flash on LMArena Agent: 3.7 → 2. (details)
Claude Sonnet 5 (high) on LMArena Agent: 7.4 → 6.3. (details)
Claude Sonnet 4.6 on LMArena Agent: 0.5 → -1. (details)
Gemini 3.8 Flash (high) on LMArena Agent: 5.5 → 4. (details)
…and 37 more.

Speed changes

GPT-6 Astra (max) output speed changed from 28 tok/s to 54 tok/s. (details)
GPT-6 Astra (max) time to first token changed from 4.63 s to 7.00 s. (details)
Claude Fable 5 output speed changed from 31 tok/s to 63 tok/s. (details)
Claude Fable 5.1 (xhigh) output speed changed from 40 tok/s to 58 tok/s. (details)
Claude Fable 5.1 (xhigh) time to first token changed from 4.77 s to 6.84 s. (details)
Claude Fable 5.1 (max) output speed changed from 40 tok/s to 67 tok/s. (details)
Claude Fable 5.1 (max) time to first token changed from 4.77 s to 6.84 s. (details)
Claude Fable 5.1 output speed changed from 40 tok/s to 67 tok/s. (details)
Claude Fable 5.1 time to first token changed from 4.77 s to 6.84 s. (details)
GPT-6 Astra output speed changed from 28 tok/s to 54 tok/s. (details)
GPT-6 Astra time to first token changed from 4.63 s to 7.00 s. (details)
Claude Fable 5.1 (high) output speed changed from 40 tok/s to 54 tok/s. (details)
Claude Fable 5.1 (high) time to first token changed from 4.77 s to 6.84 s. (details)
GPT-6 Astra (high) output speed changed from 28 tok/s to 49 tok/s. (details)
GPT-6 Astra (high) time to first token changed from 4.63 s to 7.00 s. (details)
…and 487 more.

Today's top five

More: New AI model releases this week, and the full LLM leaderboard 2026.

This digest is generated automatically from the day's data changes. Every line links to the page where you can check the numbers and their source.