Daily digest, 21 Sept 2026

BenchLeader refreshed 1519 models from 20 of 20 sources. 10 price changes, 5 new or updated benchmark results across 20 sources.

Price changes

• Qwen3.8 27B (xhigh) output price moved from $3.00/M to $2.50/M (17% down). (details)
• Ling 3.0 Flash VL input price moved from $0.06/M to $0.08/M (33% up). (details)
• Ling 3.0 Flash VL output price moved from $0.18/M to $0.22/M (22% up). (details)
• Qwen3.8 27B output price moved from $3.00/M to $2.50/M (17% down). (details)
• Qwen3.8 27B (medium) output price moved from $3.00/M to $2.50/M (17% down). (details)
• Qwen3.8 27B (low) output price moved from $3.00/M to $2.50/M (17% down). (details)
• Qwen3.8 27B (no reasoning) output price moved from $3.00/M to $2.50/M (17% down). (details)
• Granite 4.2 8B input price moved from $0.06/M to $0.10/M (67% up). (details)
• Llama 3.1 70B input price moved from $0.56/M to $0.72/M (29% up). (details)
• Llama 3.1 70B output price moved from $0.56/M to $0.72/M (29% up). (details)

New benchmark results

• FrontierMath Tiers 1–3: 1 new result, including Muse Spark 1.3 (board)
• FrontierMath Tier 4: 1 new result, including Muse Spark 1.3 (board)
• Chess Puzzles: 1 new result, including Muse Spark 1.3 (board)
• Mystery Game Puzzles: 1 new result, including Muse Spark 1.3 (board)

Score revisions

• Nova 2.0 Pro Preview (low) on AA-Omniscience: -40.5 → -46.5. (details)

Speed changes

• Claude Fable 5.1 time to first answer changed from 14.70 s to 22.53 s. (details)
• GLM 5.3 time to first token changed from 1.26 s to 1.86 s. (details)
• Qwen3.8 2.4T A95B time to first token changed from 1.87 s to 1.05 s. (details)
• Gemini 3.8 Flash time to first answer changed from 16.44 s to 22.70 s. (details)
• Claude Opus 4.8 time to first answer changed from 20.50 s to 36.74 s. (details)
• Gemini 3.5 Flash time to first answer changed from 15.12 s to 21.08 s. (details)
• Gemini 3.1 Pro time to first answer changed from 35.72 s to 63.62 s. (details)
• Claude Sonnet 5 time to first answer changed from 5.41 s to 11.94 s. (details)
• GPT-5.2 time to first answer changed from 123.31 s to 170.83 s. (details)
• Gemini 3.6 Flash time to first answer changed from 18.78 s to 25.82 s. (details)
• Kimi K2.6 time to first token changed from 1.40 s to 0.88 s. (details)
• Claude Sonnet 4.6 time to first answer changed from 81.31 s to 125.76 s. (details)
• Muse Glimmer output speed changed from 81 tok/s to 137 tok/s. (details)
• Muse Glimmer time to first token changed from 1.41 s to 0.76 s. (details)
• GPT-5 mini time to first answer changed from 86.18 s to 125.12 s. (details)
• …and 25 more.

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.