State of the leaderboard: week 40, 28 September–4 October 2026
Complete week: data as of , from 5 daily updates
The week in 6 sentences
- In week 40 of 2026 (28 September–4 October 2026), Claude Opus 5.5 (Anthropic) took the #1 spot from GPT-6 Astra on the BenchLeader Index at 69.8, 1.3 points ahead of GPT-6 Astra (68.5).
- 2 models entered the top 10 (GPT-6.1 Sol at #4 and Gemini 4 Argon at #7), pushing out Muse Spark 1.3 and GPT-5.4; 8 other models changed places.
- 10 new models arrived; the highest-ranked is GPT-6.1 Sol (OpenAI) at #4 with an index of 67.9.
- The biggest riser was GPT-6 Sol, up 0.8 points to 64.9; the biggest faller MiMo-V2.5-Pro, down 4.8 to 55.0.
- 5 models cut a list price and 10 models raised one; the largest cut was Gemma 3n E4B's input price, 100% lower.
- 613 new benchmark results were published, the most on PRBench Finance (68).
Top 10 at the end of the week
The BenchLeader Index, one row per model at its best reasoning effort, against the table at the end of the previous week. Left the top 10: Muse Spark 1.3, GPT-5.4.
| # | Model | Provider | Index | Last week | Move |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 69.8 | 6 | ↑5 |
| 2 | GPT-6 Astra | OpenAI | 68.5 | 1 | ↓1 |
| 3 | Claude Fable 5.1 | Anthropic | 68.0 | 2 | ↓1 |
| 4 | GPT-6.1 Sol | OpenAI | 67.9 | – | new |
| 5 | Claude Fable 5 | Anthropic | 66.9 | 4 | ↓1 |
| 6 | Claude Opus 5 | Anthropic | 66.7 | 3 | ↓3 |
| 7 | Gemini 4 Argon | 66.0 | – | new | |
| 8 | GPT-5.6 Sol | OpenAI | 65.9 | 5 | ↓3 |
| 9 | GPT-5.5 | OpenAI | 65.1 | 7 | ↓2 |
| 10 | Muse Spark | Meta | 65.1 | 9 | ↓1 |
New models and their first results
- Daybreak Blue (OpenAI): not enough independent results for an index yet
- Daybreak Red (OpenAI): not enough independent results for an index yet
- Claude Sonnet 5.5 (Anthropic): #19 on the index at 63.7. First results: 1786 on Arena WebDev; 1819.1 on ALE-Bench; 95.6% on GPQA Diamond; 88.8% on FrontierMath Tiers 1–3.
- Claude Opus 5.5 Range (Anthropic): not enough independent results for an index yet
- Claude Sonnet 5 Range (Anthropic): not enough independent results for an index yet
- Resonant 1 (Reson8): not enough independent results for an index yet
- Gemini 4 Argon (Google): #7 on the index at 66.0
- GPT-6.1 Sol (OpenAI): #4 on the index at 67.9. First results: 1483 on Arena Text; 1509 on Arena Hard Prompts; 1542 on Arena Coding; 1285 on Arena Vision.
- Mercury Voice (Inception): not enough independent results for an index yet
- Ling 3.1 Flash (Ant Group): not enough independent results for an index yet
Biggest movers on the index
| Model | Provider | 28 Sep 2026 | 4 Oct 2026 | Change |
|---|---|---|---|---|
| MiMo-V2.5-Pro | Xiaomi | 59.8 | 55.0 | -4.8 |
| MiMo-V2.6-Flash | Xiaomi | 59.3 | 57.9 | -1.4 |
| Gemini 3.1 Pro | 61.3 | 60.3 | -1.0 | |
| Claude Opus 5.5 | Anthropic | 70.7 | 69.8 | -0.9 |
| GPT-6 Astra | OpenAI | 69.3 | 68.5 | -0.8 |
| Claude Fable 5.1 | Anthropic | 68.8 | 68.0 | -0.8 |
| Claude Opus 5 | Anthropic | 67.5 | 66.7 | -0.8 |
| GPT-6 Sol | OpenAI | 64.1 | 64.9 | +0.8 |
| GPT-5.6 Terra | OpenAI | 63.2 | 62.5 | -0.7 |
| DeepSeek V4.1 Flash | DeepSeek | 61.3 | 60.6 | -0.7 |
Price changes
Cheaper (5)
- Gemma 3n E4B input price moved from $0.060/M to $0.00/M (100% down).
- Magistral Small input price moved from $0.50/M to $0.15/M (70% down).
- Qwen3.7 Plus output price moved from $3.00/M to $1.60/M (47% down).
- Nemotron 3.5 Lightning input price moved from $0.070/M to $0.060/M (14% down).
- R1 Distill Llama 70B input price moved from $0.80/M to $0.70/M (13% down).
More expensive (10)
- Qwen3.5 Flash input price moved from $0.029/M to $0.10/M (245% up).
- MiMo-V2.5-Pro output price moved from $1.04/M to $3.00/M (187% up).
- DeepSeek V4 Pro (max) output price moved from $0.87/M to $1.98/M (128% up).
- GPT-5.6 Sol Pro (max) input price moved from $2.00/M to $4.00/M (100% up).
- Solar Mini4 input price moved from $0.050/M to $0.10/M (100% up).
- gpt-oss-20b (high) output price moved from $0.20/M to $0.30/M (50% up).
- MiniMax-M1 input price moved from $0.40/M to $0.55/M (38% up).
- Muse Glimmer output price moved from $1.20/M to $1.50/M (25% up).
- Qwen3.7 Flash (no reasoning) output price moved from $0.12/M to $0.13/M (10% up).
- Qwen3.8 27B (xhigh) input price moved from $0.42/M to $0.45/M (7% up).
List prices per million tokens; each model's full price history is on its page and the LLM API pricing tracker.
Other weeks
← Week 41, 5–11 Oct 2026 · Week 39, 21–27 Sep 2026 → · The weekly review on the blog
- Week 45, 2–8 Nov 2026
- Week 44, 26 Oct–1 Nov 2026
- Week 43, 19–25 Oct 2026
- Week 42, 12–18 Oct 2026
- Week 41, 5–11 Oct 2026
- Week 40, 28 Sep–4 Oct 2026
- Week 39, 21–27 Sep 2026
- Week 38, 14–20 Sep 2026
- Week 37, 7–13 Sep 2026
Benchmark categories: AI agent benchmarks · AI coding benchmarks · Reasoning and math benchmarks · Multimodal AI benchmarks · New AI model releases
How this is built: the BenchLeader Index is recomputed with every update from independent benchmark publishers (see the methodology), and a maker's own launch numbers are never used. Past ranks are rebuilt from the daily index history of the models listed today.
Cite as: BenchLeader, “AI leaderboard week 40, 28 Sep–4 Oct 2026: who was #1”, data as of 4 Oct 2026, https://www.benchleader.com/state/2026-w40.