GPT-5.6 Sol Keenable added (21 Oct 2026)

BenchLeader refreshed 1609 models from 24 of 24 data feeds (benchmark publishers, aggregators and price lists). 1 new model, 5 top-10 rank changes, 78 price changes, 497 new or updated benchmark results across 24 data feeds.

New models

GPT-5.6 Sol Keenable (max) (OpenAI) is now listed; no independent results yet. (details)

Movement in the top 10

Claude Fable 5.1 (max) moved from #5 to #4 in the BenchLeader Index. (details)
GPT-6 Astra (max) moved from #4 to #5 in the BenchLeader Index. (details)
Claude Fable 5.1 (xhigh) moved from #10 to #9 in the BenchLeader Index. (details)
Claude Opus 5.5 (high) moved from #12 to #10 in the BenchLeader Index. (details)
Claude Sonnet 5.5 (max) moved from #9 to #13 in the BenchLeader Index. (details)

Price changes

Gemini 3.7 Flash (high) input price moved from $0.75/M to $0.38/M (49% down). (details)
Gemini 3.7 Flash (high) output price moved from $3.75/M to $1.88/M (50% down). (details)
DeepSeek V4.1 Flash (max) cost per task moved from 0.5 to 0.4 (18% down). (details)
Step 5 Preview cached input price moved from $0.048/M to $0.10/M (118% up). (details)
DeepSeek V4 Flash (max) cost per task moved from 0.2 to 0.1 (15% down). (details)
Kimi K2.5 (thinking) output price moved from $2.75/M to $3.00/M (9% up). (details)
DeepSeek V4 Pro (high) input price moved from $0.44/M to $0.66/M (50% up). (details)
DeepSeek V4 Pro (high) output price moved from $0.87/M to $1.98/M (128% up). (details)
DeepSeek V4 Flash (high) input price moved from $0.11/M to $0.15/M (36% up). (details)
DeepSeek V4 Flash (high) output price moved from $0.23/M to $0.60/M (161% up). (details)
Qwen3.8 27B (xhigh) cached input price moved from $0.45/M to $0.14/M (69% down). (details)
Nemotron 3 Ultra 550B A55B (thinking) cost per task moved from 0.6 to 0.5 (19% down). (details)
Nemotron 3 Ultra 550B A55B (thinking) cached input price moved from $0.15/M to $0.20/M (33% up). (details)
Nemotron 3 Ultra 550B A55B input price moved from $0.50/M to $0.60/M (20% up). (details)
Nemotron 3 Ultra 550B A55B output price moved from $2.50/M to $3.00/M (20% up). (details)
Nemotron 3 Ultra 550B A55B cached input price moved from $0.15/M to $0.20/M (33% up). (details)
Qwen3.5 397B A17B (thinking) cached input price moved from $0.22/M to $0.35/M (59% up). (details)
Qwen3.8 27B input price moved from $0.42/M to $0.45/M (7% up). (details)
Qwen3.8 27B cached input price moved from $0.45/M to $0.14/M (69% down). (details)
Qwen3.6 27B (thinking) cached input price moved from $0.30/M to $0.37/M (24% up). (details)
Qwen3.8 27B (medium) cached input price moved from $0.45/M to $0.14/M (69% down). (details)
Nemotron 3.5 Super VL 120B A12B input price moved from $0.00/M to $0.57/M (57000000000% up). (details)
Nemotron 3.5 Super VL 120B A12B output price moved from $0.00/M to $1.44/M (144000000000% up). (details)
Qwen3.5-35B-A3B (thinking) cached input price moved from $0.050/M to $0.15/M (200% up). (details)
Qwen3.8 27B (low) cached input price moved from $0.45/M to $0.14/M (69% down). (details)
Gemma 4 31B (thinking) input price moved from $0.14/M to $0.20/M (43% up). (details)
Qwen3.5 397B A17B cached input price moved from $0.22/M to $0.35/M (59% up). (details)
Step 3.7 Flash cached input price moved from $0.037/M to $0.040/M (9% up). (details)
Beam 1.0 cost per task moved from 0.6 to 0.5 (9% down). (details)
Qwen3.5 397B A17B (no reasoning) cached input price moved from $0.22/M to $0.35/M (59% up). (details)
…and 48 more.

New benchmark results

MLCR: 91 new results, including Gemini 4 Argon, Claude Opus 5.5, GPT-6 Astra… (board)
Arena Maths: 20 new results, including Muse Wonder, DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B… (board)
Arena Text: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Hard Prompts: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Coding: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Creative Writing: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Instruction Following: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Multi-turn: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
Arena Longer Queries: 19 new results, including DeepSeek V4 Pro, Nemotron 3 Ultra 550B A55B, Kimi K2.5… (board)
FrontierCode 1.1 (AA): 16 new results, including Muse Wonder Full, GPT-6.1 Sol, GPT-6.1 Sol… (board)
Terminal-Bench-Science 0.1 (AA): 13 new results, including GPT-6.1 Sol, Muse Spark 1.3, GPT-6.1 Sol… (board)
ARC-AGI-1: 7 new results, including Claude Sonnet 5.5, Claude Sonnet 5.5, Claude Sonnet 5.5… (board)

Score revisions

Claude Opus 5.5 (max) on MLCR: 66.7% → 27.8%. (details)
Claude Opus 5.5 (max) on Arena Agent: 20.4 → 20.9. (details)
Claude Fable 5.1 (max) on MLCR: 71.1% → 15.6%. (details)
GPT-6 Astra (max) on MLCR: 35% → 41.7%. (details)
GPT-6.1 Sol (max) on MLCR: 33.9% → 38.3%. (details)
GPT-6.1 Sol (max) on Arena Coding: 1540 → 1534. (details)
GPT-6.1 Sol (max) on Arena Maths: 1494 → 1512. (details)
GPT-6.1 Sol (max) on Arena Multi-turn: 1486 → 1479. (details)
GPT-6.1 Sol (max) on Arena Longer Queries: 1492 → 1487. (details)
Claude Opus 5.5 (high) on Arena Maths: 1502 → 1512. (details)
Claude Fable 5 (max) on MLCR: 64.4% → 20.6%. (details)
Claude Sonnet 5.5 (max) on MLCR: 75% → 41.7%. (details)
Claude Opus 5 (max) on MLCR: 55.6% → 11.1%. (details)
GPT-5.6 Sol (max) on MLCR: 26.1% → 30%. (details)
Claude Opus 5 (high) on MLCR: 59.4% → 11.7%. (details)
GPT-6 Astra on Epoch Capabilities Index: 166.4 → 168.8. (details)
Muse Wonder (max) on Arena Creative Writing: 1485 → 1490. (details)
Muse Wonder (max) on Arena Multi-turn: 1530 → 1522. (details)
Claude Opus 5 (xhigh) on MLCR: 58.3% → 10%. (details)
Muse Spark 1.3 (max) on MLCR: 43.3% → 13.9%. (details)
…and 148 more.

Speed changes

Claude Opus 5 time to first answer changed from 20.94 s to 13.58 s. (details)
GPT-5.5 Pro time to first token changed from 47.39 s to 4.19 s. (details)
GPT-6 Sol time to first answer changed from 15.86 s to 9.18 s. (details)
GPT-5.6 Terra time to first answer changed from 17.11 s to 9.66 s. (details)
GLM 5.3 Flash time to first token changed from 0.94 s to 0.61 s. (details)
DeepSeek V4 Pro time to first answer changed from 45.81 s to 25.76 s. (details)
DeepSeek V4 Pro response time changed from 50.84 s to 31.79 s. (details)
DeepSeek V4 Pro thinking time changed from 44.02 s to 24.11 s. (details)
Step 5 Preview time to first token changed from 3.49 s to 1.19 s. (details)
GPT-6 Luna time to first token changed from 4.16 s to 2.15 s. (details)
Claude Opus 4.5 time to first token changed from 2.32 s to 1.49 s. (details)
DeepSeek V4 Flash time to first answer changed from 105.12 s to 16.39 s. (details)
DeepSeek V4 Flash response time changed from 114.34 s to 20.16 s. (details)
DeepSeek V4 Flash thinking time changed from 103.48 s to 15.07 s. (details)
DeepSeek V4 Flash time to first token changed from 2.05 s to 1.23 s. (details)
…and 84 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.