Anthropic
Claude Opus 4.8 is Anthropic’s generally available Opus-family model for complex reasoning and agentic coding, with Claude API model ID `claude-opus-4-8`.
Claude Opus 4.8 has strong official evidence for being Anthropic’s current generally available Opus-family flagship: documented API ID, 1M-token context on supported platforms, 128k max output, image/file inputs, and broad deployment through Claude Platform, AWS, Google Cloud, and Microsoft Foundry. These support high technical and developer-experience scores, though the catalog record should replace the PDF system card as the canonical website/documentation URL. Coding evidence is unusually strong but partly self-reported. Anthropic’s system card reports 88.6% on SWE-bench Verified, and BenchmarkList mirrors that result while marking provenance as self-reported. DeepSWE/DataCurve lists claude-opus-4.8[max] at 58%±2%, second on that leaderboard, with high average cost and long runtime. LiveCodeBench was checked but the official leaderboard did not list Claude Opus 4.8, so no LiveCodeBench score is credited. Pricing clarity is good from Anthropic and Artificial Analysis, but cost-effectiveness is only moderate: Opus-tier pricing remains expensive even with batch or third-party variants. Artificial Analysis reports useful speed/provider measurements, but not enough to make it a low-latency specialist. Public signal is solid from LMArena rankings and vote counts, plus Anthropic’s distribution footprint. Risk/evidence quality is good overall, with limitations from self-reported benchmark provenance, absent Terminal-Bench/Aider results, and some catalog-source metadata inconsistencies.