Daily digest, 20 Sept 2026

BenchLeader refreshed 1519 models from 20 of 20 sources. 9 price changes, 220 new or updated benchmark results across 20 sources.

Price changes

• Qwen3.8 27B (xhigh) output price moved from $2.55/M to $3.00/M (18% up). (details)
• Qwen3.8 27B output price moved from $2.55/M to $3.00/M (18% up). (details)
• Muse Glimmer input price moved from $0.35/M to $0.30/M (14% down). (details)
• Muse Glimmer output price moved from $1.50/M to $1.20/M (20% down). (details)
• Qwen3.8 27B (medium) output price moved from $2.55/M to $3.00/M (18% up). (details)
• Qwen3.8 27B (low) output price moved from $2.55/M to $3.00/M (18% up). (details)
• Muse Glimmer (high) input price moved from $0.35/M to $0.30/M (14% down). (details)
• Muse Glimmer (high) output price moved from $1.50/M to $1.20/M (20% down). (details)
• Qwen3.8 27B (no reasoning) output price moved from $2.55/M to $3.00/M (18% up). (details)

Score revisions

• Claude Fable 5.1 (high) on GDPval (AA): 57.5% → 55.9%. (details)
• GPT-6 Astra (high) on GDPval (AA): 51.5% → 49.2%. (details)
• GPT-6 Astra (max) on GDPval (AA): 54% → 52.1%. (details)
• Claude Fable 5.1 (xhigh) on GDPval (AA): 62.3% → 61.1%. (details)
• Claude Fable 5.1 (thinking) on GDPval (AA): 61.2% → 61.7%. (details)
• GPT-6 Astra (xhigh) on GDPval (AA): 52.8% → 50.8%. (details)
• Claude Fable 5 (thinking) on GDPval (AA): 56.6% → 54.8%. (details)
• Claude Opus 5 (high) on GDPval (AA): 56.5% → 54.1%. (details)
• Claude Opus 5 (xhigh) on GDPval (AA): 60.4% → 58.8%. (details)
• Claude Opus 5 (max) on GDPval (AA): 61.8% → 60.4%. (details)
• GPT-6 Astra (medium) on GDPval (AA): 50% → 48.4%. (details)
• Claude Fable 5.1 (medium) on GDPval (AA): 54% → 51.8%. (details)
• Muse Spark 1.3 (max) on GDPval (AA): 60.2% → 58.7%. (details)
• GPT-5.6 Sol (xhigh) on GDPval (AA): 54.2% → 52.4%. (details)
• GPT-6 Astra (low) on GDPval (AA): 46% → 43.3%. (details)
• GPT-5.6 Sol (high) on GDPval (AA): 51.2% → 49%. (details)
• Muse Spark 1.3 (xhigh) on GDPval (AA): 58.1% → 56.4%. (details)
• GPT-5.5 (xhigh) on GDPval (AA): 44.8% → 41.8%. (details)
• Claude Fable 5.1 (low) on GDPval (AA): 50.2% → 47.5%. (details)
• Kimi K3 (max) on GDPval (AA): 52.4% → 51.2%. (details)
• …and 200 more.

Speed changes

• Claude Fable 5.1 time to first answer changed from 106.60 s to 157.09 s. (details)
• Claude Opus 5 time to first answer changed from 37.52 s to 21.36 s. (details)
• Qwen3.8 2.4T A95B time to first token changed from 3.20 s to 1.87 s. (details)
• GPT-5.4 Pro output speed changed from 1 tok/s to 2 tok/s. (details)
• Gemini 3.5 Flash time to first answer changed from 15.00 s to 20.38 s. (details)
• GPT-5.1-Codex output speed changed from 49 tok/s to 25 tok/s. (details)
• Muse Glimmer output speed changed from 25 tok/s to 81 tok/s. (details)
• Qwen-Plus time to first token changed from 0.28 s to 0.47 s. (details)
• MiniMax-M1 output speed changed from 65 tok/s to 24 tok/s. (details)
• Qwen3.6 27B output speed changed from 33 tok/s to 46 tok/s. (details)
• Qwen2.5 72B time to first token changed from 0.66 s to 0.42 s. (details)
• Llama 3.1 Nemotron 70B Instruct output speed changed from 33 tok/s to 75 tok/s. (details)
• Nemotron Nano 12B v2 VL output speed changed from 49 tok/s to 68 tok/s. (details)
• Gemma 2 27B time to first token changed from 0.74 s to 1.20 s. (details)
• Lyria 3 Pro time to first token changed from 1.08 s to 1.49 s. (details)

Today's top five

1. Claude Fable 5.1 — 72.0
2. GPT-6 Astra — 71.9
3. GPT-6 Astra — 71.8
4. Claude Fable 5.1 — 71.6
5. Claude Fable 5.1 — 71.3

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.