AI benchmark results and model performance
Individual benchmark scores plotted by date.
| Organisation | Model | Reported | Top Score | Info | Self Reported | Source |
|---|---|---|---|---|---|---|
| Qwen 2.5 Coder 32B Instruct | - | 5.92% | inferred high-confidence family alias from qwen2.5-vl-32b (score=0.4641; benches=28) | Yes | Source | |
| Qwen 2.5 VL 32B Instruct | 24 Mar 2025 | 5.92% | - | Yes | Source | |
| Qwen 2.5 VL 3B Instruct | - | 5.92% | inferred high-confidence family alias from qwen2.5-vl-32b (score=0.4914; benches=28) | Yes | Source | |
| Qwen 2.5 Math 72B | - | 8.83% | inferred high-confidence family alias from qwen2.5-vl-72b (score=0.4700; benches=30) | Yes | Source | |
| Qwen 2.5 Math PRM 72B | - | 8.83% | inferred family alias from qwen2.5-vl-72b (score=0.4050; benches=30) | Yes | Source | |
| Qwen 2.5 Math RM 72B | - | 8.83% | inferred family alias from qwen2.5-vl-72b (score=0.4092; benches=30) | Yes | Source | |
| Qwen 2.5 VL 72B Instruct | 26 Jan 2025 | 8.83% | - | Yes | Source | |
| Qwen 3 VL 4B Instruct | - | 26.20% | - | Yes | - | |
| Qwen 3 VL 30B A3B Instruct | - | 30.30% | - | Yes | - | |
| Qwen 3 VL 30B A3B Thinking | - | 30.60% | - | Yes | - | |
| Qwen 3 4B Instruct 2507 | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48) | Yes | - | |
| Qwen 3 4B SafeRL | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48) | Yes | - | |
| Qwen 3 4B Thinking 2507 | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48) | Yes | - | |
| Qwen 3 Embedding 4B | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48) | Yes | - | |
| Qwen 3 Guard Gen 4B | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3400; benches=48) | Yes | - | |
| Qwen 3 Guard Stream 4B | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3371; benches=48) | Yes | - | |
| Qwen 3 Reranker 4B | - | 31.40% | inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48) | Yes | - | |
| Qwen 3 VL 4B Thinking | - | 31.40% | - | Yes | - | |
| Qwen 3 VL 32B Instruct | - | 32.60% | - | Yes | - | |
| Qwen 3 8B | - | 33.90% | inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.4600; benches=50) | Yes | - | |
| Qwen 3 Guard Stream 8B | - | 33.90% | inferred family alias from qwen3-vl-8b-thinking (score=0.3371; benches=50) | Yes | - | |
| Qwen 3 VL 8B Instruct | - | 33.90% | - | Yes | - | |
| Qwen 3 VL 8B Thinking | - | 33.90% | - | Yes | - | |
| Qwen 3 VL Embedding 8B | - | 33.90% | inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5232; benches=50) | Yes | - | |
| Qwen 3 VL Reranker 8B | - | 33.90% | inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5275; benches=50) | Yes | - | |
| Qwen 3 VL 235B A22B Thinking | - | 38.10% | - | Yes | - | |
| GPT 5.4 Nano | 17 Mar 2026 | 39% | Verified | Yes | Source | |
| Qwen 3 VL 32B Thinking | - | 41% | - | Yes | - | |
| Claude Haiku 4.5 | 15 Oct 2025 | 50.70% | inferred alias from claude-haiku-4-5-20251001 | Yes | Source | |
| Qwen 3.5 35B A3B | 24 Feb 2026 | 54.50% | - | Yes | Source | |
| Qwen 3.5 27B | 24 Feb 2026 | 56.20% | - | Yes | Source | |
| Qwen 3.5 Flash | 23 Feb 2026 | 56.20% | inferred family alias from qwen3.5-27b (score=0.4147; benches=81) | Yes | Source | |
| Qwen 3.5 122B A10B | 24 Feb 2026 | 58% | - | Yes | Source | |
| GLM 5V Turbo | 01 Apr 2026 | 62.30% | - | Yes | Source | |
| Qwen 3.6 Plus | 01 Apr 2026 | 62.50% | - | Yes | Source | |
| GPT 5.3 Codex | 05 Feb 2026 | 64.70% | OSWorld-Verified; xhigh reasoning | Yes | Source | |
| Claude Opus 4.5 | 24 Nov 2025 | 66.26% | Pass@1; Avg@5, 64k Thinking | Yes | Source | |
| Qwen 3 VL 235B A22B Instruct | - | 66.70% | - | Yes | - | |
| MiniMax M3 | 01 Jun 2026 | 70% | Verified | Yes | Source | |
| GPT 5.4 Mini | 17 Mar 2026 | 72.10% | Verified | Yes | Source | |
| Claude Sonnet 4.6 | 17 Feb 2026 | 72.50% | OSWorld-Verified | Yes | Source | |
| Claude Opus 4.6 | 05 Feb 2026 | 72.70% | OSWorld-Verified | Yes | Source | |
| GPT 5 Pro | 07 Aug 2025 | 75% | inferred family alias from gpt-5.4 (score=0.4083; benches=19) | Yes | Source | |
| GPT 5 Search API | 14 Oct 2025 | 75% | inferred family alias from gpt-5.4 (score=0.3050; benches=19) | Yes | Source | |
| GPT 5.4 | 05 Mar 2026 | 75% | Verified | Yes | Source | |
| Claude Opus 4.7 | 16 Apr 2026 | 78% | OSWorld-Verified | Yes | Source | |
| GPT 5.5 | 23 Apr 2026 | 78.70% | Verified | Yes | Source | |
| Claude Mythos Preview | 07 Apr 2026 | 79.60% | Verified | Yes | Source | |
| Muse Spark 1.1 | 09 Jul 2026 | 80.80% | OSWorld-Verified; agentic computer use | Yes | Source | |
| Claude Sonnet 5 | 30 Jun 2026 | 81.20% | OSWorld-Verified; pass@1 averaged over five runs; 361 tasks, 100 steps; adaptive thinking at max effort | Yes | Source | |
| Claude Opus 4.8 | 28 May 2026 | 83.40% | OSWorld-Verified; updated Anthropic evaluation setup | Yes | Source | |
| Claude Fable 5 | 09 Jun 2026 | 85% | OSWorld-Verified; higher of Mythos 5 / Fable 5 in Anthropic's launch table | Yes | Source | |
| Claude Mythos 5 | 09 Jun 2026 | 85% | OSWorld-Verified; higher of Mythos 5 / Fable 5 in Anthropic's launch table | Yes | Source |
