Mistral Large 4.0 OSS added, GLM 4.5V price cut (3 Nov 2026)
Mistral Large 4.0 OSS joins the board; GLM 4.5V got cheaper. 1 new model, 81 price changes, 34 new benchmark results. Daily digest for 3 Nov 2026.
BenchLeader refreshed 1683 models from 27 of 27 data feeds (benchmark publishers, aggregators and price lists). 1 new model, 81 price changes, 34 new or updated benchmark results across 27 data feeds.
New models
- Mistral Large 4.0 OSS (high) (Mistral AI) appeared with 5 benchmark results, entering the BenchLeader Index at #336. (details)
Price changes
- GLM 5.2 (max) cached input price moved from $0.26/M to $0.30/M (15% up). (details)
- GLM-5 (thinking) cached input price moved from $0.20/M to $0.15/M (25% down). (details)
- GLM 5.3 (max) cached input price moved from $0.26/M to $0.30/M (15% up). (details)
- GLM 5.3 Flash cached input price moved from $0.030/M to $0.034/M (15% up). (details)
- GLM 5.1 input price moved from $1.40/M to $0.90/M (36% down). (details)
- GLM 5.1 output price moved from $4.40/M to $3.59/M (19% down). (details)
- GLM 5.1 cached input price moved from $0.26/M to $0.19/M (25% down). (details)
- GLM 5V Turbo input price moved from $1.20/M to $0.75/M (38% down). (details)
- GLM 5V Turbo output price moved from $4.00/M to $3.29/M (18% down). (details)
- GLM 5V Turbo cached input price moved from $0.24/M to $0.18/M (25% down). (details)
- GLM-5 input price moved from $1.00/M to $0.60/M (40% down). (details)
- GLM-5 output price moved from $3.20/M to $2.69/M (16% down). (details)
- GLM-5 cached input price moved from $0.20/M to $0.15/M (25% down). (details)
- GLM 5.3 Flash (max) cached input price moved from $0.030/M to $0.034/M (15% up). (details)
- GLM 4.7 input price moved from $0.60/M to $0.45/M (25% down). (details)
- GLM 4.7 cached input price moved from $0.11/M to $0.090/M (19% down). (details)
- GLM 5.2 input price moved from $1.40/M to $1.20/M (15% down). (details)
- GLM 5.2 cached input price moved from $0.26/M to $0.30/M (15% up). (details)
- GLM 4.5V input price moved from $0.60/M to $0.30/M (50% down). (details)
- GLM 4.5V output price moved from $1.80/M to $0.90/M (50% down). (details)
- GLM 4.5V cached input price moved from $0.11/M to $0.060/M (46% down). (details)
- GLM 5.3 (low) cached input price moved from $0.26/M to $0.30/M (15% up). (details)
- GLM-4.5-Air cached input price moved from $0.030/M to $0.024/M (20% down). (details)
- GLM 5.2 (no reasoning) cached input price moved from $0.26/M to $0.30/M (15% up). (details)
- DeepSeek V4 Pro 0424 (max) input price moved from $1.41/M to $1.31/M (7% down). (details)
- GLM 4.5V (no reasoning) cached input price moved from $0.11/M to $0.060/M (46% down). (details)
- GLM 4.5V (thinking) cached input price moved from $0.11/M to $0.060/M (46% down). (details)
- GLM 4.6V (no reasoning) cached input price moved from $0.050/M to $0.030/M (40% down). (details)
- GLM 4.6V (thinking) cached input price moved from $0.050/M to $0.030/M (40% down). (details)
- GLM 4.6V input price moved from $0.30/M to $0.15/M (50% down). (details)
- …and 51 more.
New benchmark results
- ARC-AGI-3: 1 new result, including DeepSeek V4.1 Flash 1017 (board)
- BrowseComp: 1 new result, including gpt-oss-20b (board)
- Terminal-Bench-Science 0.1 (AA): 1 new result, including Apodex 1.1 (board)
- Arena Agent: 1 new result, including Spark X2.6 (board)
Score revisions
- Claude Opus 5.5 (max) on Arena Agent: 20.9 → 20. (details)
- Claude Opus 5.5 (max) on Coding-agent share: 53.8% → 54.4%. (details)
- Gemini 4 Carbon (high) on Arena Agent: 16 → 16.6. (details)
- GPT-6 Astra (max) on Arena Agent: 13 → 13.7. (details)
- Claude Opus 5.5 (xhigh) on Coding-agent share: 53.8% → 54.4%. (details)
- Claude Opus 5.5 (high) on Coding-agent share: 53.8% → 54.4%. (details)
- Claude Sonnet 5.5 (max) on Arena Agent: 13.8 → 14.6. (details)
- GPT-6 Sol (max) on Arena Agent: 8 → 6.9. (details)
- Claude Opus 5.5 on Coding-agent share: 53.8% → 54.4%. (details)
- Muse Wonder Full (max) on Arena Agent: 10.1 → 9.6. (details)
- Muse Wonder (max) on Arena Agent: 3.1 → 2.5. (details)
- Claude Opus 5.5 (medium) on Coding-agent share: 53.8% → 54.4%. (details)
- GLM 5.2 (max) on Arena Agent: 1.3 → 0.7. (details)
- Claude Opus 5.5 (low) on Coding-agent share: 53.8% → 54.4%. (details)
- DeepSeek V4.1 Flash 1017 (max) on Arena Agent: 2.7 → 2.1. (details)
- DeepSeek V4.1 Flash 1017 (max) on OpenRouter token share: 24.8% → 25.5%. (details)
- Grok 4.7 (xhigh) on AA Intelligence Index v5.0: 34.3 → 35.3. (details)
- Grok 4.7 (xhigh) on Terminal-Bench-Science 0.1 (AA): 4.8% → 11.4%. (details)
- Claude Haiku 5.5 (high) on Arena Agent: 1.4 → 0.7. (details)
- MiMo-V2.6-Flash on Arena Agent: 1 → 0.4. (details)
- …and 10 more.
Speed changes
- GPT-6 Astra time to first token changed from 6.16 s to 3.34 s. (details)
- GPT-6.1 Sol time to first token changed from 6.06 s to 3.34 s. (details)
- GPT-6.1 Sol output speed changed from 85 tok/s to 42 tok/s. (details)
- Kimi K3 time to first token changed from 1.80 s to 0.86 s. (details)
- GPT-5.6 Terra time to first token changed from 1.38 s to 1.88 s. (details)
- Claude Sonnet 5 time to first answer changed from 10.29 s to 16.19 s. (details)
- Kimi K3 output speed changed from 37 tok/s to 87 tok/s. (details)
- GPT-5-Codex time to first answer changed from 11.90 s to 17.22 s. (details)
- GPT-5.6 Luna time to first token changed from 3.18 s to 2.02 s. (details)
- Grok 4.20 Multi-Agent time to first token changed from 15.45 s to 7.65 s. (details)
- DeepSeek V4 Pro time to first token changed from 0.73 s to 1.35 s. (details)
- Grok 4.7 time to first answer changed from 43.19 s to 66.05 s. (details)
- Grok 4.7 response time changed from 49.74 s to 72.67 s. (details)
- Claude Opus 5 time to first answer changed from 7.67 s to 10.38 s. (details)
- DeepSeek V4 Flash output speed changed from 77 tok/s to 49 tok/s. (details)
- …and 51 more.
Today's top five
- Claude Fable 5.2 (max) — 72.7
- Gemini 4 Carbon (high) — 68.9
- Claude Opus 5.5 (max) — 68.4
- GPT-6 Astra (max) — 67.3
- Claude Fable 5.1 (max) — 66.7
More: New AI model releases this week, and the full LLM leaderboard 2026.
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.