FUNDING · MODELS · #805
Anthropic releases Opus 5.5 with improved performance and lower pricing
Anthropic announced Opus 5.5, a faster and cheaper follow-up to Opus 5 that the company says outperforms the larger Fable model on many coding and knowledge-work benchmarks. Output token pricing has been reduced (now $20 per mTok vs $25 for Opus 5), the model communicates with less jargon and front-loads key information, and Sonnet 5.5 and Haiku 5.5 are slated to follow soon; the release is subject to the same safety safeguards as Fable and underwent external evaluation.
KEY POINTS
- Anthropic announced Opus 5.5, a faster and cheaper follow-up to Opus 5 that the company says outperforms the larger Fable model on many coding and knowledge-work benchmarks.
- Output token pricing has been reduced (now $20 per mTok vs $25 for Opus 5), the model communicates with less jargon and front-loads key information, and Sonnet 5.5 and Haiku 5.5 are slated to follow soon; the release is subject to the same safety safeguards as Fable and underwent external evaluation.
- This matters because Opus 5.5 advances frontier model performance while lowering cost and compute needs, shifting commercial and safety dynamics among leading LLM developers.
DECISION BRIEF
CONFIDENCE
WHAT CHANGED
Anthropic released Claude Opus 5.5. The company says Opus 5.5 outperforms its larger Fable model on many coding and knowledge-work benchmarks, runs faster, and lowers compute costs. Output token pricing was reduced to $20 per million output tokens (Opus 5 was $25); input token price reported at $4 per million (down from $5). Cache-read costs were reported to be reduced substantially. Opus 5.5 communicates with less jargon and front-loads key information. The model is available on Amazon Bedrock and Claude Platform on AWS. Anthropic applied safeguards and safety classifiers similar to Fable 5.1 (including rerouting some cybersecurity and biology requests to less-capable models) and ran pre-release external evaluations (METR, Frontier Design). Sonnet 5.5 and Haiku 5.5 are announced as coming soon.
WHY NOW
Opus 5.5 combines reported frontier-model performance improvements with lower per-token and lower operating costs and faster runtime; those changes can affect commercial unit economics for agentic coding and long-form knowledge work and alter how Anthropic applies safety controls at higher capability levels. The release also follows Anthropic’s public shift toward ‘pacing the frontier,’ making safety design and external evaluation a stated priority.
WHO IS AFFECTED
Directly affected: Anthropic cloud customers and developers using Claude (including via Amazon Bedrock), teams running agentic coding and long-running knowledge-work tasks, early adopters evaluating communication/usability improvements, and external evaluators/safety teams that assessed the release (METR, Frontier Design). Industries called out as fits: software development and knowledge work; reporting also mentions cybersecurity and biology-related safeguards.
CONFIRMED
All items below are explicitly stated in the supplied sources: - Anthropic released Claude Opus 5.5. (947, 955, 948, 949) - Anthropic and reporting say Opus 5.5 outperforms or matches Fable 5.1 on many coding/knowledge-work benchmarks (phrased as “outpaces” or “matches … on most tasks”). (947, 949, 948) - Output token pricing: $20 per million output tokens (Opus 5 was $25). Input token pricing reported at $4 per million (down from $5). Cache-read costs reported materially reduced (e.g., 60% reduction reported). (947, 949, 955) - Anthropic reports Opus 5.5 runs faster and uses fewer tokens per task (claims of >30% faster outputs and ~40% lower operating cost in company reporting). (949, 947, 955) - Opus 5.5 aims to communicate more clearly: less jargon, front-loads key info, surfaces what it did/what it found/what it needs; includes adaptive thinking/“effort” control. (947, 955, 949, 1033) - Opus 5.5 is available on Amazon Bedrock and Claude Platform on AWS. (955) - The release uses safety classifiers and safeguards similar to Fable 5.1 (including routing some cybersecurity requests to Opus 4.8 and some biology requests to Opus 5) and was evaluated pre-release by outside organizations including METR and Frontier Design. (947, 948, 955) - Anthropic announced Claude Sonnet 5.5 and Haiku 5.5 are expected “in the coming weeks.” (947, 949, 948) - Anthropic stated the release is the first since CEO Dario Amodei’s statement about pacing frontier progress.
UNCERTAIN
Missing, limited, or conflicting evidence in the supplied excerpts: - Independent confirmation: Many performance and cost claims are presented as Anthropic’s results or company claims; the excerpts note external partners evaluated the model but do not provide independent, detailed benchmark results or datasets for public review. (947, 949, 948) - Scope and generalizability: It is unclear how the reported cost and speed improvements translate across diverse, real-world workloads and customer configurations; sources report both token-price cuts (20%) and larger total operating-cost reductions (~40%) without a public, standardized accounting that reconciles them. (947, 949, 955) - Safeguards effectiveness: Sources state Opus 5.5 uses stricter safeguards, classifiers, and request rerouting, but do not provide quantitative measures (e.g., refusal rates, false positives/negatives) or operational telemetry showing how those systems perform in practice. (948, 955, 947) - Relative benchmark language varies: TechCrunch says Opus 5.5 “outpaces the larger Fable model on many benchmarks,” while Decoder frames it as “matches Fable 5.1 on most tasks” and provides cost comparisons to other vendors; the excerpts do not supply the raw benchmark data to resolve phrasing differences. (947, 949) - Timing/details for Sonnet 5.5/Haiku 5.5: announced as coming soon but no release dates or specs in the excerpts. (947, 949, 948) - User experience claims (reduced “Claudish” writing): Anthropic staff and early testers report clearer writing, but broader user studies or systematic human-evaluation results are not provided in the excerpts.
WHAT TO WATCH
Observable signals to monitor (concrete, source-linked): - Publication of independent benchmark results or raw data from third parties (e.g., FrontierCode, Terminal-Bench, CursorBench) or independent labs that corroborate Anthropic’s claims. (949, 947) - Release or public reports from METR, Frontier Design, or other external evaluators detailing their Opus 5.5 assessments. (947, 948) - Availability and timing announcements for Sonnet 5.5 and Haiku 5.5 and their published specs/performance claims. (947, 949, 948) - AWS Bedrock adoption signals: documentation updates, pricing/throughput telemetry, or customer case studies showing Opus 5.5 usage patterns and cost-per-task in production. (955) - Empirical reports on safeguard behavior: observable refusal rates, examples of rerouted cybersecurity/biology queries, or incident reports showing whether new classifiers prevent or fail to prevent risky outputs. (948, 955, 947) - User feedback and evaluations reporting changes in “Claudish” writing (surveys, comparative human-evaluation studies, or published A/B tests).
WHY IT MATTERS
This matters because Opus 5.5 advances frontier model performance while lowering cost and compute needs, shifting commercial and safety dynamics among leading LLM developers.
EVIDENCE MAP
4Editorial claims linked to specific sources, with support, contradiction and context shown separately.
Anthropic released Opus 5.5 as a faster, lower-priced successor to Opus 5.
SUPPORTEDChecked 2026-09-24 · 1 supporting
Reported output-token pricing is per million tokens, down from for Opus 5.
SUPPORTEDChecked 2026-09-24 · 1 supporting
Opus 5.5 is available through Amazon Bedrock.
SUPPORTEDChecked 2026-09-24 · 1 supporting
Independent coverage says the release adds stricter cybersecurity safeguards.
UNRESOLVEDChecked 2026-09-24 · 0 supporting
SOURCES & TIMELINE
4Anthropic’s newest model, Opus 5.5, was released on Tuesday, setting a new state-of-the-art in coding and knowledge work performance. Notably, the release outpaces the larger Fable model in many benchmarks, and succeeded in a number of informal tasks that Fable failed to complete. In the announcement, Anthropic called it “the strongest-performing model we’ve tested to date.” Opus 5.5 is also significantly cheaper t…
Today, we’re excited to announce the availability of Claude Opus 5.5 on Amazon Bedrock and Claude Platform on AWS , the first of the Claude 5.5 model family. Claude Opus 5.5 is Anthropic’s most capable Opus model suitable for agentic coding, knowledge work, and long-running tasks. This post covers Claude Opus 5.5’s improvements, practical guidance, and how to start building with the model on Amazon Bedrock. Accordi…
Claude Opus 5.5 comes with improvements to certain behaviors, like attempting to escape testing environments. Anthropic says its new Claude Opus 5.5 model comes with stronger safeguards in the wake of recent rogue AI hacking incidents. In an announcement on Tuesday, Anthropic says Opus 5.5 comes with improvements to certain risky behaviors, including attempts to escape the company’s testing sandbox. It’s the first …
Anthropic is launching Claude Opus 5.5, the first model in a new family. The company says it delivers Claude Fable 5.1-level performance while costing significantly less and running faster than its predecessor. According to Anthropic, Opus 5.5 matches Claude Fable 5.1 "on most tasks" while costing about 40 percent less to run than Opus 5. Claude Sonnet 5.5 and Haiku 5.5 are expected in the coming weeks, with similar…
AI models are improving fast at math, code, and reasoning. Their writing quality, though, has stalled or even gotten worse. Anthropic employee Jackson Kernion , who works on Claude fine-tuning, recently explained why Opus 4.6 was the last good "writing model" from Anthropic. The answer is partly predictable: newer models have been optimized for math and code. But they've also been trained to produce technical expla…