#swe-bench

Claude Fable 5 tops SWE-bench Verified at 95%, but 99 of 100 results are self-reported and the scaffold gap can exceed 28 points. How to read the numbers.

We compare Claude Opus 4.8 and GPT-5.5 on coding, agents, reasoning, and real cost — including where GPT-5.5 still wins and which model fits which job.

How to read AI model leaderboards without being fooled by benchmark contamination, eval gaming, and cherry-picked MMLU, GPQA, and SWE-bench scores.

Five-way comparison — Claude Code, Cursor, Codex Desktop, Replit Agent 3, Devin. Pricing, agent autonomy, MCP, eval scores, and reference workloads.

Six production-tested GPT-5.5 Pro coding workflows — refactor, review, debug, test-gen, migration, codebase Q&A — with cost, latency, and success-rate data.

Claude Opus 4.7 scores 64.3% on SWE-bench Pro with 2576px vision, xhigh effort, and same Opus 4.6 pricing. Full benchmark and migration guide.

Cursor Composer 2 beats Opus 4.6 on coding benchmarks at 90% lower cost. Built on Kimi K2.5. GPT-5.4 still leads. Full benchmark comparison guide.

GPT-5.4 Mini launches for free-tier ChatGPT users with 54.38% SWE-Bench Pro performance, only 3 points behind full GPT-5.4. 2x faster guide.

NVIDIA releases Nemotron 3 Super 120B with 60.47% SWE-Bench Verified and 2.2x throughput. Open-source coding model for enterprise AI agent deployments.

Gemini 3.1 Pro vs Claude Opus 4.6 vs GPT-5.3-Codex for agentic coding. SWE-Bench, Terminal-Bench, LiveCodeBench, and pricing comparison with recommendations.

MiniMax M2.5 scores 80.2% SWE-Bench Verified and costs 1/10th of competitors. Complete guide to features, benchmarks, pricing, API access, and model comparison.

Xiaomi's MiMo-V2-Flash: 309B open-weight MoE running 150 tok/s with 73.4% SWE-Bench. Deploy the fastest open-source coding model with this guide.

Master Devstral 2 (72.2% SWE-bench) and Mistral Vibe CLI. Open-weight coding models that run locally. Complete autonomous agent guide.

Master Claude Opus 4.5: 80.9% SWE-bench, Memory Tool, self-improving agents. Complete guide with pricing and API integration.

MiniMax M2 achieves 69.4 on SWE-bench at 8% of Claude's cost. Complete guide to China's open-source AI model for agents, coding & multimodal apps.