Daily digest, 16 Sept 2026

BenchLeader refreshed 1366 models from 20 of 20 sources. 6 new models, 2 top-10 rank changes, 70 price changes, 23 new or updated benchmark results across 20 sources.

New models

Qwen3 8 (Alibaba) appeared with 4 benchmark results. (details)
Solar Pro 4 (Upstage) appeared with 1 benchmark result. (details)
Claude Sonnet 4 (APAC) (Amazon) is now listed at $6.00/M blended; no independent results yet. (details)
Claude Sonnet 4 (EU) (Amazon) is now listed at $6.00/M blended; no independent results yet. (details)
Claude Sonnet 4 (Global) (Amazon) is now listed at $6.00/M blended; no independent results yet. (details)
Claude Sonnet 4 (US) (Amazon) is now listed at $6.00/M blended; no independent results yet. (details)

Movement in the top 10

Claude Fable 5 moved from #4 to #3 in the BenchLeader Index. (details)
GPT-6 Astra (high) moved from #3 to #4 in the BenchLeader Index. (details)

Price changes

MiniMax-M1 input price moved from $0.55/M to $0.40/M (27% down). (details)
AU Anthropic Claude Opus 4.6 input price moved from $16.50/M to $5.50/M (67% down). (details)
AU Anthropic Claude Opus 4.6 output price moved from $82.50/M to $27.50/M (67% down). (details)
AU Anthropic Claude Opus 4.6 cached input price moved from $1.65/M to $0.55/M (67% down). (details)
Claude Fable 5 (US) input price moved from $10.00/M to $11.00/M (10% up). (details)
Claude Fable 5 (US) output price moved from $50.00/M to $55.00/M (10% up). (details)
Claude Fable 5 (US) cached input price moved from $1.00/M to $1.10/M (10% up). (details)
Claude Haiku 4.5 (AU) input price moved from $1.00/M to $1.10/M (10% up). (details)
Claude Haiku 4.5 (AU) output price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Haiku 4.5 (AU) cached input price moved from $0.10/M to $0.11/M (10% up). (details)
Claude Haiku 4.5 (JP) input price moved from $1.00/M to $1.10/M (10% up). (details)
Claude Haiku 4.5 (JP) output price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Haiku 4.5 (JP) cached input price moved from $0.10/M to $0.11/M (10% up). (details)
Claude Haiku 4.5 (US) input price moved from $1.00/M to $1.10/M (10% up). (details)
Claude Haiku 4.5 (US) output price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Haiku 4.5 (US) cached input price moved from $0.10/M to $0.11/M (10% up). (details)
Claude Opus 4.5 (US) input price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Opus 4.5 (US) output price moved from $25.00/M to $27.50/M (10% up). (details)
Claude Opus 4.5 (US) cached input price moved from $0.50/M to $0.55/M (10% up). (details)
Claude Opus 4.6 (US) input price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Opus 4.6 (US) output price moved from $25.00/M to $27.50/M (10% up). (details)
Claude Opus 4.6 (US) cached input price moved from $0.50/M to $0.55/M (10% up). (details)
Claude Opus 4.7 (JP) input price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Opus 4.7 (JP) output price moved from $25.00/M to $27.50/M (10% up). (details)
Claude Opus 4.7 (JP) cached input price moved from $0.50/M to $0.55/M (10% up). (details)
Claude Opus 4.7 (US) input price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Opus 4.7 (US) output price moved from $25.00/M to $27.50/M (10% up). (details)
Claude Opus 4.7 (US) cached input price moved from $0.50/M to $0.55/M (10% up). (details)
Claude Opus 4.8 (AU) input price moved from $5.00/M to $5.50/M (10% up). (details)
Claude Opus 4.8 (AU) output price moved from $25.00/M to $27.50/M (10% up). (details)
…and 40 more.

New benchmark results

LMArena Agent: 2 new results, including Gemini 3.7 Flash, Gemini 3.8 Flash (board)
SciCode (AA): 2 new results, including Gemma 4 26B A4B, GPT-5.6 Luna (board)

Score revisions

Claude Opus 5 (max) on LMArena Agent: 11.1 → 10.2. (details)
Qwen3 8 (max) on AA Intelligence Index: 40.3 → 45.4. (details)
Qwen3 8 (max) on AA-LCR: 78.3% → 80.3%. (details)
Qwen3 8 (max) on SciCode (AA): 53.2% → 52.1%. (details)
Qwen3 8 (max) on AA-Omniscience: 3.4 → 12. (details)
Grok 4.5 on LMArena Agent: 3.4 → 2.9. (details)
Grok 4.6 (xhigh) on LMArena Agent: 2.6 → 2. (details)
Claude Sonnet 5 (high) on LMArena Agent: 5.4 → 6. (details)
Qwen3.7 Plus on LMArena Agent: -5.5 → -6.1. (details)
DeepSeek V4 Pro on LMArena Agent: -0.1 → -0.7. (details)
GPT-5.6 Luna (xhigh) on LMArena Agent: 0.2 → -0.4. (details)
Gemini 3.6 Flash (high) on LMArena Agent: -5.4 → -5.9. (details)
Inkling Small on LMArena Agent: -7.8 → -9.2. (details)
Gemini 3.5 Flash Lite on LMArena Agent: -14.2 → -15.3. (details)
GPT-5.6 Terra (no reasoning) on AA Intelligence Index: 22.3 → 21.2. (details)
Mistral Medium 3.5 on LMArena Agent: -10.3 → -10.9. (details)
North Mini Code on AA Intelligence Index: 12.8 → 9.9. (details)
GPT-5.6 Luna (no reasoning) on AA Intelligence Index: 16.8 → 16.1. (details)
Mistral Medium 3 on AA-LCR: 31.3% → 42.7%. (details)

Speed changes

Claude Opus 4.7 time to first answer changed from 17.93 s to 24.31 s. (details)
Claude Opus 5 time to first answer changed from 3.59 s to 5.24 s. (details)
GPT-5.3-Codex time to first answer changed from 40.73 s to 56.75 s. (details)
Grok 4.6 time to first answer changed from 41.74 s to 26.38 s. (details)
o3 time to first answer changed from 4.06 s to 6.73 s. (details)
Claude Sonnet 5 time to first answer changed from 2.76 s to 4.09 s. (details)
Claude Opus 4.5 time to first answer changed from 7.50 s to 15.27 s. (details)
Step 3.7 Flash time to first token changed from 2.40 s to 3.46 s. (details)
GPT-5.4 Nano time to first answer changed from 101.05 s to 61.54 s. (details)
Nemotron 3 Super time to first token changed from 1.90 s to 2.74 s. (details)
MiMo-V2.5-Pro time to first answer changed from 6.73 s to 4.19 s. (details)
MiMo-V2.5-Pro time to first token changed from 6.73 s to 4.19 s. (details)
Nemotron 3 Nano 30B A3B output speed changed from 223 tok/s to 142 tok/s. (details)
Nemotron 3 Nano 30B A3B time to first answer changed from 10.14 s to 15.37 s. (details)
Gemini 2.5 Flash time to first answer changed from 15.23 s to 24.59 s. (details)
…and 13 more.

Today's top five

Claude Fable 5.1 — 70.5
Claude Fable 5.1 — 70.3
Claude Fable 5 — 70.1
GPT-6 Astra — 70.1
Claude Fable 5.1 — 69.9

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.