Anthropic launches Claude Opus 5.5, claims Fable-5.1-level performance at 40% lower cost than Opus 5
2026-09-22
Anthropic released Claude Opus 5.5 (claude-opus-5-5) on September 22, 2026, positioned for long-running agentic coding and knowledge work, with a 1M-token context window by default, 128k max output tokens, and always-on adaptive thinking (thinking cannot be disabled). Pricing is $4/$20 per MTok input/output versus Opus 5's $5/$25 — a 40% cost reduction Anthropic claims comes with performance 'at the level of Claude Fable 5.1 on most work.' It is available via the Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, and Microsoft Foundry; computer use on this model requires the newer computer_toolset_20260801 toolset.
Significance 4: A new flagship model release with a claimed frontier-parity capability jump at a substantial (40%) price cut is a tracked-axis-moving development, but the performance claim is lab-reported and capped at 4 pending independent verification.
Implications · machine-drafted, not owner judgment
If the parity-with-Fable-5.1 claim holds up independently, Anthropic has effectively cut the price of its top agentic-coding tier by 40% while closing the gap to its most capable model, which pressures OpenAI and Google on price-per-quality for coding-agent workloads and could reopen the coding-agent and daily-driver verdicts currently held by Opus 4.8. TechCrunch frames this as part of a 'model drop week' alongside a same-day OpenAI GPT-6 update, suggesting the frontier labs are now racing to match each other's release cadence within hours, not weeks.
- Independent benchmark reproduction of the Fable-5.1-parity claim (SWE-bench Verified, Terminal-Bench, AA Intelligence Index)
- Whether real-world agentic sessions on Opus 5.5 show the same recovery-from-bad-states behavior that won Opus 4.8 the coding-agent pick
Sources
- primary Claude Platform release notes retrieved 2026-09-26
- primary Introducing Claude Opus 5.5 retrieved 2026-09-26
- press Meta's Muse just stole the AI spotlight from OpenAI and Anthropic retrieved 2026-09-26
Affected axes
Coding · Hard reasoning · Cost · Agentic / tool use · Long context