Daily digest, 26 Sept 2026

BenchLeader refreshed 1555 models from 22 of 22 sources. 114 new or updated benchmark results across 22 sources.

New benchmark results

• LMArena Maths: 10 new results, including GPT-6 Astra, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Text: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Hard Prompts: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Coding: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Creative Writing: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Instruction Following: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Multi-turn: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• LMArena Longer Queries: 8 new results, including Claude Opus 5.5, GPT-6 Sol, Claude Fable 5… (board)
• PRBench Legal: 2 new results, including Grok 4.7, Claude Opus 5.5 (board)
• LMArena Agent: 1 new result, including GPT-6 Sol (board)
• SciCode (AA): 1 new result, including Qwen3.5-9B (board)
• FORTRESS: 1 new result, including Claude Opus 5.5 (board)

Score revisions

• GPT-6 Astra (max) on LMArena Hard Prompts: 1497 → 1502. (details)
• GPT-6 Astra (max) on LMArena Creative Writing: 1461 → 1453. (details)
• GPT-6 Astra (max) on LMArena Instruction Following: 1461 → 1473. (details)
• GPT-6 Astra (max) on LMArena Longer Queries: 1481 → 1488. (details)
• Claude Fable 5.1 (max) on LMArena Coding: 1519 → 1529. (details)
• Claude Opus 5 (max) on LMArena Maths: 1523 → 1515. (details)
• Muse Spark 1.3 (max) on LMArena Maths: 1496 → 1505. (details)
• Muse Spark 1.3 (max) on LMArena Longer Queries: 1506 → 1498. (details)
• GPT-5.6 Sol (xhigh) on LMArena Maths: 1488 → 1494. (details)
• GLM 5.3 (max) on LMArena Maths: 1504 → 1497. (details)
• GLM 5.3 (max) on LMArena Creative Writing: 1462 → 1456. (details)
• GLM 5.3 (max) on LMArena Multi-turn: 1493 → 1481. (details)
• Gemini 3.8 Flash (high) on LMArena Creative Writing: 1496 → 1482. (details)
• Gemini 3.8 Flash (high) on LMArena Multi-turn: 1513 → 1505. (details)
• Grok 4.6 (high) on LMArena Creative Writing: 1451 → 1445. (details)
• Gemini 3.7 Flash (high) on LMArena Maths: 1522 → 1507. (details)
• GLM 5.2 (max) on LMArena Maths: 1480 → 1487. (details)
• GLM 5.2 (max) on LMArena Multi-turn: 1469 → 1475. (details)
• GLM 5.3 Flash on LMArena Maths: 1513 → 1501. (details)
• Grok 4.5 on LMArena Coding: 1519 → 1513. (details)
• …and 23 more.

Speed changes

• GPT-6 Astra time to first answer changed from 66.16 s to 38.57 s. (details)
• GPT-6 Astra response time changed from 75.41 s to 47.11 s. (details)
• Claude Fable 5 time to first answer changed from 106.90 s to 153.30 s. (details)
• Claude Fable 5 response time changed from 114.40 s to 160.11 s. (details)
• GPT-6 Sol output speed changed from 114 tok/s to 55 tok/s. (details)
• GPT-5.3-Codex time to first answer changed from 51.23 s to 71.64 s. (details)
• GPT-5.3-Codex response time changed from 55.39 s to 75.80 s. (details)
• Gemini 3.8 Flash time to first answer changed from 16.41 s to 22.57 s. (details)
• Qwen3.8 2.4T A95B time to first token changed from 1.04 s to 1.69 s. (details)
• Gemini 3.5 Flash time to first answer changed from 13.22 s to 20.98 s. (details)
• Gemini 3.5 Flash response time changed from 15.69 s to 23.14 s. (details)
• Claude Opus 4.8 time to first answer changed from 24.17 s to 15.30 s. (details)
• Grok 4.7 time to first answer changed from 1.10 s to 2.81 s. (details)
• GPT-5.6 Luna time to first answer changed from 17.27 s to 7.36 s. (details)
• GPT-5.6 Luna response time changed from 21.59 s to 11.42 s. (details)
• …and 15 more.

Today's top five

1. Claude Opus 5.5 — 72.0
2. Claude Fable 5.1 — 70.8
3. GPT-6 Astra — 70.2
4. Claude Opus 5.5 — 70.2
5. Gemini 4 Argon — 70.0

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.