Claude Fable 5.249.1%
2.0 / Cognition native agentseffort: xhigh · Cognition evaluation; 132 tasks, mean@5; Claude Code; highest-performing effort is xhigh, not max. · Tools enabled
Anthropic ↗Best LLMs for mergeable code changes
FrontierCodeTests whether an agent can produce software changes that meet functional and review requirements, helping you assess how much implementation work you could delegate before human code review. Scores cover 132 tasks and depend on the agent setup and effort level.
Updated: Data as of:
| Model / creator | Version | Reported score |
|---|---|---|
 Claude Fable 5.2Anthropic · Proprietary | 2.0 / independent mini-swe-agent | 49.128% |
| Claude Fable 5.2Anthropic · Proprietary | 2.0 / Cognition native agents | 49.1% |
| Claude Opus 5.5Anthropic · Proprietary | 2.0 / Cognition native agents | 45.5% |
 GPT-6 AstraOpenAI · Proprietary | 2.0 / Cognition native agents | 42.8% |
 Gemini 4 CarbonGoogle · Proprietary | 2.0 / Cognition native agents | 41.1% |
| Claude Fable 5.1Anthropic · Proprietary | 2.0 / Cognition native agents | 32.4% |
View all 10 published reports
Claude Opus 5.545.5%
2.0 / Cognition native agentseffort: max · November 9 launch-table comparison; highest available effort unless explicitly noted. Preserve the benchmark track and model configuration; competitor figures are as reported by Anthropic. · Tools enabled
Anthropic ↗GPT-6 Astra42.8%
2.0 / Cognition native agentseffort: max · November 9 launch-table comparison; highest available effort unless explicitly noted. Preserve the benchmark track and model configuration; competitor figures are as reported by Anthropic. · Tools enabled
Anthropic ↗Gemini 4 Carbon41.1%
2.0 / Cognition native agentseffort: medium · November 9 launch-table comparison; highest available effort unless explicitly noted. Preserve the benchmark track and model configuration; competitor figures are as reported by Anthropic. · Tools enabled
Anthropic ↗Claude Fable 5.132.4%
2.0 / Cognition native agentseffort: xhigh · November 9 launch-table comparison; highest available effort unless explicitly noted. Preserve the benchmark track and model configuration; competitor figures are as reported by Anthropic. · Tools enabled
Anthropic ↗Claude Fable 5.249.128%
2.0 / independent mini-swe-agenteffort: high · Claude Fable 5.2 (High, Default Fallback); exact independent record; implementation per archived benchmarking methodology. · Tools enabled
Claude Fable 5.248.76%
2.0 / independent mini-swe-agenteffort: max · Claude Fable 5.2 (Max, Default Fallback); exact independent record; implementation per archived benchmarking methodology. · Tools enabled
Claude Fable 5.247.581%
2.0 / independent mini-swe-agenteffort: xhigh · Claude Fable 5.2 (Xhigh, Default Fallback); exact independent record; implementation per archived benchmarking methodology. · Tools enabled
Claude Fable 5.241.298%
2.0 / independent mini-swe-agenteffort: medium · Claude Fable 5.2 (Medium, Default Fallback); exact independent record; implementation per archived benchmarking methodology. · Tools enabled
Claude Fable 5.238.44%
2.0 / independent mini-swe-agenteffort: low · Claude Fable 5.2 (Low, Default Fallback); exact independent record; implementation per archived benchmarking methodology. · Tools enabled