Kimi K3.1 Pro added (23 Oct 2026)

BenchLeader refreshed 1617 models from 24 of 24 data feeds (benchmark publishers, aggregators and price lists). 7 new models, 130 price changes, 350 new or updated benchmark results across 24 data feeds.

New models

Kimi K3.1 Pro (Moonshot AI) appeared with 6 benchmark results, entering the BenchLeader Index at #37. (details)
Nemotron 3.5 Bolt E4B 4GB (thinking) (NVIDIA) appeared with 1 benchmark result. (details)
Nemotron Super VL 3.5 NVFP4 (NVIDIA) appeared with 1 benchmark result. (details)
Smaug Nemotron 3.5 NVFP4 (NVIDIA) appeared with 1 benchmark result. (details)
Spark X2.6 (Synlan) appeared with 1 benchmark result. (details)
Kimi K3.1 Pro Highspeed (Moonshot AI) is now listed at $8.00/M blended; no independent results yet. (details)
Qwen3.8 Max Internal (Alibaba) is now listed; no independent results yet. (details)

Price changes

Qwen3.8 Max (max) cached input price moved from $0.25/M to $0.22/M (11% down). (details)
Muse Spark Open (xhigh) cached input price moved from $1.25/M to $0.15/M (88% down). (details)
DeepSeek V4.1 Flash (max) cost per task moved from $0.40/M to $0.48/M (20% up). (details)
Qwen3.7 (max) cached input price moved from $0.50/M to $0.33/M (34% down). (details)
Qwen3.6 Plus cached input price moved from $0.050/M to $0.028/M (45% down). (details)
Qwen3.7 Plus cached input price moved from $0.040/M to $0.055/M (38% up). (details)
Qwen3.5 Plus (thinking) input price moved from $0.40/M to $0.12/M (71% down). (details)
Qwen3.5 Plus (thinking) output price moved from $2.40/M to $0.69/M (71% down). (details)
MiMo-V2.5-Pro (thinking) input price moved from $0.44/M to $0.41/M (7% down). (details)
MiMo-V2.5-Pro (thinking) output price moved from $0.87/M to $1.02/M (17% up). (details)
MiMo-V2.5-Pro (thinking) cost per task moved from $0.054/M to $0.14/M (158% up). (details)
Mistral Large 4 cached input price moved from $0.070/M to $0.14/M (100% up). (details)
Qwen3.8 27B (xhigh) cached input price moved from $0.14/M to $0.10/M (28% down). (details)
Quasar 438B (max) input price moved from $0.60/M to $0.11/M (82% down). (details)
Quasar 438B (max) output price moved from $1.80/M to $0.33/M (82% down). (details)
Nemotron 3 Ultra 550B A55B (thinking) cached input price moved from $0.20/M to $0.12/M (40% down). (details)
Nemotron 3 Ultra 550B A55B cached input price moved from $0.20/M to $0.12/M (40% down). (details)
Qwen3.5 397B A17B (thinking) cached input price moved from $0.35/M to $0.33/M (6% down). (details)
Inkling Small cached input price moved from $0.10/M to $0.12/M (16% up). (details)
Qwen3.8 27B cached input price moved from $0.14/M to $0.10/M (28% down). (details)
MiMo-V2.5 input price moved from $0.14/M to $0.13/M (7% down). (details)
MiMo-V2.5 cost per task moved from $0.019/M to $0.038/M (99% up). (details)
Ling 3.0 Flash Fin input price moved from $0.080/M to $0.075/M (6% down). (details)
Qwen3.8 27B (medium) cached input price moved from $0.14/M to $0.10/M (28% down). (details)
Qwen3.8 27B (low) cached input price moved from $0.14/M to $0.10/M (28% down). (details)
Qwen3.5 Flash input price moved from $0.10/M to $0.029/M (71% down). (details)
Qwen3.5 Flash output price moved from $0.40/M to $0.29/M (28% down). (details)
Qwen3.5 397B A17B cached input price moved from $0.35/M to $0.33/M (6% down). (details)
Qwen3 32B input price moved from $0.70/M to $0.29/M (59% down). (details)
Qwen3 32B output price moved from $2.80/M to $1.15/M (59% down). (details)
…and 100 more.

New benchmark results

Harvey's Legal Agent Benchmark v1: 77 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Excel Modeling Benchmark v1.1: 28 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Finance Agent v2.1: 26 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Vals Finance Index: 26 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Vals RSI Index: 24 new results, including Claude Opus 5.5, Gemini 4 Argon, Muse Wonder Full… (board)
Code Migration v1.1: 24 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Vals Science Index: 22 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
SWE-bench Multimodal: 20 new results, including Gemini 4 Argon, Muse Wonder Full, GPT-6 Astra… (board)
Vals Router Index: 19 new results, including GPT-6 Astra, Gemini 4 Argon, MiMo-V2.6-Pro… (board)
SWE-bench Verified (any scaffold): 11 new results, including Claude Opus 4.5, GPT-5, Claude Sonnet 4.5… (board)
Robo-bench: 7 new results, including Claude Opus 5.5, Gemini 4 Argon, Muse Wonder Full… (board)
ClaimBench: 5 new results, including GPT-6.1 Sol, Claude Opus 5.5, GPT-6 Luna… (board)

Score revisions

Gemini 4 Argon (high) on Arena Text: 1524 → 1517. (details)
Gemini 4 Argon (high) on Arena Hard Prompts: 1549 → 1540. (details)
Gemini 4 Argon (high) on Arena Coding: 1559 → 1552. (details)
Gemini 4 Argon (high) on Arena WebDev: 1678 → 1673. (details)
Gemini 4 Argon (high) on Arena Instruction Following: 1525 → 1519. (details)
Gemini 4 Argon (high) on Arena Multi-turn: 1552 → 1544. (details)
Gemini 4 Argon (high) on Arena Longer Queries: 1543 → 1535. (details)
Muse Spark 1.3 (max) on Arena Maths: 1504 → 1510. (details)
Gemini 3 Pro on SWE-bench Verified (any scaffold): 74.2% → 77.4%. (details)
MiMo-V2.6-Flash on Arena Maths: 1471 → 1464. (details)
Mistral Large 4 on Arena Multi-turn: 1437 → 1443. (details)
Nemotron 3.5 Super VL 120B A12B on Arena Maths: 1407 → 1416. (details)
Claude Opus 4 on SWE-bench Verified (any scaffold): 67.6% → 73.2%. (details)
Claude Sonnet 4 on SWE-bench Verified (any scaffold): 74.8% → 76.8%. (details)
Kimi K2 on SWE-bench Verified (any scaffold): 53.4% → 65.4%. (details)
Claude 3.7 Sonnet on SWE-bench Verified (any scaffold): 52.8% → 66.4%. (details)
Qwen3 Coder 480B on SWE-bench Verified (any scaffold): 55.4% → 69.6%. (details)
GPT-4o on SWE-bench Verified (any scaffold): 21.6% → 38.8%. (details)
Gemini 2.0 Flash on SWE-bench Verified (any scaffold): 13.5% → 52.2%. (details)
Qwen2.5 Coder 32B Instruct on SWE-bench Verified (any scaffold): 38% → 47%. (details)
…and 8 more.

Speed changes

Claude Fable 5.1 time to first answer changed from 257.00 s to 164.48 s. (details)
Claude Opus 5.5 time to first answer changed from 35.38 s to 21.50 s. (details)
Claude Opus 5 time to first answer changed from 44.14 s to 20.24 s. (details)
Claude Opus 5 response time changed from 53.92 s to 29.14 s. (details)
GPT-5.6 Sol time to first answer changed from 90.20 s to 56.61 s. (details)
GPT-5.6 Sol response time changed from 96.39 s to 62.65 s. (details)
Muse Spark 1.3 time to first token changed from 2.01 s to 2.93 s. (details)
GPT-5.5 Pro output speed changed from 19 tok/s to 26 tok/s. (details)
GPT-5.5 Pro time to first token changed from 11.41 s to 4.35 s. (details)
GLM 5.3 time to first token changed from 0.83 s to 1.13 s. (details)
GPT-5.4 Pro output speed changed from 7 tok/s to 10 tok/s. (details)
Qwen3.8 2.4T A95B time to first token changed from 6.05 s to 1.07 s. (details)
GPT-5.6 Terra time to first answer changed from 14.85 s to 7.49 s. (details)
GPT-5.6 Terra response time changed from 19.50 s to 12.17 s. (details)
Muse Spark 1.1 time to first token changed from 1.16 s to 1.61 s. (details)
…and 99 more.

Today's top five

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.