gpt-5.5
84.7%±2.1%
82.6%CI86.8%
aiqbench · LLM benchmarks
Benchmarks
- Sources
- 8
- Entries tracked
- 22,308
- Last sync
- 15 min15 minutes ago
All sources syncing normally
- Arena · Text· 11 hours ago
- Arena · WebDev· 11 hours ago
- Arena · Vision· 11 hours ago
- Arena · Agent· 11 hours ago
- GPQA Diamond· 15 minutes ago
- SWE-bench· 15 minutes ago
- Terminal-Bench· 15 minutes ago
- ARC-AGI-2· 15 minutes ago
Filter by org:
Released:
gpt-5.4-2026-03-05
81.8%±2.0%
79.8%CI83.8%
claude-opus-4-7
80.2%±2.1%
78.1%CI82.3%
One grid, every column. Pick axes in the Axes panel — one axis ranks like a single board, several blend into a weighted composite, all of them read as the index. Rank is assigned before any filter and never renumbers.
| Model | Org | ± CI | Swing | Add to compare | |||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | gpt-5.5 | OpenAI | Sep 2026 | $5 / $30 | 84.7% | 82.6%–86.8% | #9–34 | ||
| 2 | gpt-5.4-2026-03-05 | OpenAI | Sep 2026 | $2.50 / $15 | 81.8% | 79.8%–83.8% | #12–29 | ||
| 3 | claude-opus-4-7 | Anthropic | Sep 2026 | $5 / $25 | 80.2% | 78.1%–82.3% | #1–18 | ||
| 4 | gemini-3.1-pro-preview | Google DeepMind | Sep 2026 | $2 / $12 | 80.2% | 77.6%–82.8% | #9–36 | ||
| 5 | claude-opus-4-6 | Anthropic | Sep 2026 | $5 / $25 | 79.8% | 78.2%–81.4% | #2–25 | ||
| 6 | gpt-5.3-codex | OpenAI | Oct 2026 | $1.75 / $14 | 78.4% | 76.2%–80.6% | #28–236 | ||
| 7 | gemini-3-pro-preview | Google DeepMind | — | — | 69.4% | 67.3%–71.5% | #19–120 | ||
| 8 | gpt-5.2-codex | OpenAI | Oct 2026 | $1.75 / $14 | 66.5% | 63.5%–69.6% | #37–370 | ||
| 9 | gpt-5.2-2025-12-11 | OpenAI | Dec 2025 | $1.75 / $14 | 64.9% | 62.2%–67.7% | #34–92 | ||
| 10 | gemini-3-flash-preview | Google DeepMind | — | $0.50 / $3 | 64.3% | 61.5%–67.1% | #27–99 | ||
| 11 | claude-opus-4-5-20251101 | Anthropic | Nov 2025 | $5 / $25 | 63.1% | 60.4%–65.8% | #20–62 | ||
| 12 | gpt-5.1-codex-mini | OpenAI | Oct 2026 | $0.25 / $2 | 61.6% | 59.7%–63.5% | #54–425 | ||
| 13 | gpt-5.1-codex-max | OpenAI | Oct 2026 | $1.25 / $10 | 60.4% | 57.8%–63.1% | #52–375 | ||
| 14 | grok-4-20 | xAI | — | $1.25 / $2.50 | 57.3% | — | #44–259 | ||
| 15 | claude-sonnet-4-6 | Anthropic | Sep 2026 | $3 / $15 | 53.4% | 50.6%–56.3% | #38–77 | ||
| 16 | glm-5 | Z.ai (Zhipu AI) | Feb 2026 | $1 / $3.20 | 52.4% | 49.8%–55.0% | #57–153 | ||
| 17 | gpt-5-2025-08-07 | OpenAI | Aug 2025 | $1.25 / $10 | 49.6% | 46.7%–52.5% | #64–149 | ||
| 18 | gpt-5.1-2025-11-13 | OpenAI | Nov 2025 | $1.25 / $10 | 47.6% | 44.8%–50.4% | #45–242 | ||
| 19 | claude-sonnet-4-5-20250929 | Anthropic | Sep 2025 | $3 / $15 | 46.5% | 44.2%–48.9% | #66–232 | ||
| 20 | MiniMax-M2.7 | MiniMax | Apr 2026 | $0.30 / $1.20 | 45.1% | — | #67–241 | ||
| 21 | gpt-5-codex | OpenAI | — | — | 44.3% | 41.6%–47.0% | #17–142 | ||
| 22 | kimi-k2.5 | Moonshot | Jan 2026 | $0.50 / $2.50 | 43.2% | 40.3%–46.1% | #70–446 | ||
| 23 | MiniMax-M2.5 | MiniMax | Feb 2026 | $0.27 / $0.95 | 42.7% | 39.9%–45.5% | #82–458 | ||
| 24 | deepseek/deepseek-v3.2 | DeepSeek | — | — | 39.6% | 36.8%–42.3% | #76–585 | ||
| 25 | claude-opus-4-1-20250805 | Anthropic | Aug 2025 | $15 / $75 | 38.0% | 35.4%–40.6% | #73–404 | ||
| 26 | MiniMax-M2.1 | MiniMax | Dec 2025 | $0.30 / $1.20 | 36.6% | 33.7%–39.5% | #34–201 | ||
| 27 | kimi-k2-thinking | Moonshot | Nov 2025 | $0.60 / $2.50 | 35.7% | 32.9%–38.6% | #35–209 | ||
| 28 | claude-haiku-4-5-20251001 | Anthropic | Oct 2025 | $1 / $5 | 35.5% | 32.6%–38.4% | #117–611 | ||
| 29 | gpt-5-mini-2025-08-07 | OpenAI | Aug 2025 | $0.25 / $2 | 34.8% | 32.1%–37.5% | #126–343 | ||
| 30 | glm-4.7 | Z.ai (Zhipu AI) | Dec 2025 | $0.54 / $1.98 | 33.4% | 30.5%–36.2% | #82–189 | ||
| 31 | gemini-2.5-pro | Google DeepMind | Jun 2025 | $1.25 / $10 | 32.6% | 29.6%–35.7% | #60–432 | ||
| 32 | MiniMax-M2 | MiniMax | Oct 2025 | $0.26 / $1.02 | 30.0% | 27.3%–32.8% | #135–415 | ||
| 33 | Kimi-K2-Instruct | Moonshot | — | — | 27.8% | 25.3%–30.3% | #108–566 | ||
| 34 | grok-4-0709 | xAI | Jul 2025 | — | 27.2% | 24.1%–30.3% | #95–276 | ||
| 35 | Qwen3-Coder-480B-A35B-Instruct | Alibaba | Jul 2025 | — | 27.2% | 24.6%–29.8% | #153–508 |
Showing Terminal-Bench exactly as the source published it — its own score and confidence interval, no blending. Select a second axis to build a weighted composite.