Updated daily at 05:00 UTC
Benchmark
APEX-Agents
Measures: An agent works inside simulated professional 'worlds' (files plus workplace apps exposed as tools) to finish long-horizon, cross-application tasks written by investment-banking analysts, management consultants and corporate lawyers; an LLM judge grades each output against expert-written binary rubric criteria, and Pass@1 is the share of tasks meeting every criterion on one attempt.
- Maintainer
- Mercor
- Introduced
- 2026
- Status
- Active
- Saturation
- Headroom
- Typical provenance
- Independent
- In scoring set
- Yes
Leaderboard
Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.
| Model | Pass@1 (all rubric criteria met) | Mean score (share of rubric criteria passed) | Provenance | Source | Evaluated |
|---|---|---|---|---|---|
| Claude Sonnet 5.5 Anthropic · Max effort | 75.5% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Opus 5.5 Anthropic · Max effort | 73.5% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Fable 5.1 Anthropic · Unknown effort | 68.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| Gemini 3.7 Flash Google DeepMind · Unknown effort | 67.8% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Opus 5 Anthropic · Max effort | 65.8% | — | Source not stated | Epoch AI via Epoch AI | — |
| Grok 4.6 SpaceXAI (xAI) · Unknown effort | 65.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-6 Astra OpenAI · Unknown effort | 64.7% | — | Source not stated | Epoch AI via Epoch AI | — |
| Gemini 3.8 Flash Google DeepMind · Unknown effort | 64.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Fable 5 | 63.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Fable 5.1 Anthropic · High effort | 59.7% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-5.6 Terra OpenAI · Max effort | 58.2% | — | Source not stated | Epoch AI via Epoch AI | — |
| Muse Spark 1.3 Meta · Unknown effort | 57.8% | — | Source not stated | Epoch AI via Epoch AI | — |
| GLM-5.3 Z.ai (Zhipu AI) · Unknown effort | 56.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| Grok 4.5 SpaceXAI (xAI) · Unknown effort | 56.2% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-5.5 OpenAI · Unknown effort | 55.1% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Sonnet 5 Anthropic · Unknown effort | 54.5% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-6 Sol OpenAI · Max effort | 54.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| GLM-5.3-Flash Z.ai (Zhipu AI) · Unknown effort | 52.8% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-5.4 OpenAI · Unknown effort | 52.4% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-5.6 Sol OpenAI · Promax effort | 51.4% | — | Source not stated | Epoch AI via Epoch AI | — |
| Kimi K3 Moonshot AI · Unknown effort | 50.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Opus 4.7 Anthropic · Max effort | 49.2% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Opus 4.8 Anthropic · Max effort | 48.9% | — | Source not stated | Epoch AI via Epoch AI | — |
| DeepSeek-V4-Pro DeepSeek · Unknown effort | 47.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| Gemini 3.6 Flash Google DeepMind · Unknown effort | 46.9% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Opus 4.6 Anthropic · Max effort | 46.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Sonnet 5.5 Anthropic · Medium effort | 44.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| GPT-6 Luna OpenAI · Max effort | 44.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| Claude Sonnet 4.6 Anthropic · High effort | 43.0% | — | Source not stated | Epoch AI via Epoch AI | — |
| GLM-5.1 | 40.9% | — | Source not stated | Epoch AI via Epoch AI | — |
| MiniMax-M3 | 37.7% | — | Source not stated | Epoch AI via Epoch AI | — |
| Kimi K2.7 Code | 37.6% | — | Source not stated | Epoch AI via Epoch AI | — |
| Muse Spark 1.2 | 36.4% | — | Source not stated | Epoch AI via Epoch AI | — |
| Gemini 3.1 Pro | 35.3% | — | Source not stated | Epoch AI via Epoch AI | — |
| Inkling | 33.8% | — | Source not stated | Epoch AI via Epoch AI | — |
| Gemini 3.5 Flash Google DeepMind · Unknown effort | 27.5% | — | Source not stated | Epoch AI via Epoch AI | — |
| Qwen3.5-397B-A17B | 24.9% | — | Source not stated | Epoch AI via Epoch AI | — |
| gpt-oss-120b | 4.4% | — | Source not stated | Epoch AI via Epoch AI | — |
- Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.