Daily digest, 4 Oct 2026

BenchLeader refreshed 1550 models from 23 of 23 sources. 61 new or updated benchmark results across 23 sources.

New benchmark results

• PRBench Finance: 34 new results, including GPT-5.6 Sol, Claude Fable 5, GPT-6 Astra… (board)
• AudioMC: 27 new results, including Gemini 3.8 Flash, Inkling Small, Gemini 2.5 Flash… (board)

Speed changes

• Claude Sonnet 5 time to first token changed from 2.63 s to 1.68 s. (details)
• GPT-5 nano time to first token changed from 1.09 s to 1.54 s. (details)
• Ling 3.0 Flash Fin time to first token changed from 0.55 s to 1.09 s. (details)
• Qwen3.6 27B time to first token changed from 0.57 s to 1.15 s. (details)
• Qwen-Plus time to first token changed from 0.51 s to 0.33 s. (details)
• Nemotron 3.5 Lightning time to first token changed from 0.51 s to 0.21 s. (details)
• Trinity Large time to first token changed from 0.18 s to 0.56 s. (details)
• Mistral Medium 3 output speed changed from 48 tok/s to 68 tok/s. (details)
• Devstral 2 output speed changed from 9 tok/s to 61 tok/s. (details)
• Gemma 3 27B time to first token changed from 0.80 s to 1.09 s. (details)
• Gemma 3 12B output speed changed from 24 tok/s to 35 tok/s. (details)
• Llama 3.2 3B Instruct output speed changed from 27 tok/s to 56 tok/s. (details)
• Qwen3 8B output speed changed from 42 tok/s to 57 tok/s. (details)
• GPT-3.5 Turbo output speed changed from 46 tok/s to 69 tok/s. (details)
• Codestral 2508 output speed changed from 42 tok/s to 64 tok/s. (details)
• …and 3 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.