Anthropic has launched Claude Opus 4.6, updating its Opus-class mannequin with modifications geared toward coding workflows and long-context use, together with a 1M token context window in beta.
The San Francisco-based AI firm mentioned that is the primary Opus-class mannequin to help 1M tokens of context.
Entry deeper business intelligence
Expertise unmatched readability with a single platform that mixes distinctive knowledge, AI, and human experience.
Anthropic mentioned Claude Opus 4.6 is on the market at the moment on claude.ai, via its API and on main cloud platforms. Builders can entry it as claude-opus-4-6 through the Claude API, the corporate mentioned.
Pricing stays $5/$25 per million tokens, based on Anthropic. For prompts exceeding 200,000 tokens, the corporate mentioned premium pricing applies at $10/$37.50 per million enter/output tokens.
Anthropic additionally mentioned US-only inference is on the market at 1.1× token pricing for workloads that should run in the US.
On mannequin capabilities, Anthropic mentioned Opus 4.6 improves on Claude Opus 4.5 in coding, with extra cautious planning, longer-running agentic process execution, extra dependable operation in bigger codebases, and stronger code evaluate and debugging meant to assist the mannequin catch its personal errors.
The corporate mentioned the identical enhancements can help different information work, together with operating monetary analyses, performing analysis, and utilizing or creating paperwork, spreadsheets and displays. Moreover, these workflows can run inside Cowork, the place Claude can multitask autonomously.
Anthropic reported benchmark outcomes throughout a number of evaluations. It mentioned Opus 4.6 achieves the best rating on Terminal-Bench 2.0, which it described as an agentic coding analysis. The Claude developer additionally mentioned the mannequin leads different frontier fashions on Humanity’s Final Examination, which it described as a multidisciplinary reasoning take a look at.
On GDPval-AA, which it described as measuring efficiency on economically useful information work duties throughout finance, authorized and different domains, Anthropic mentioned Opus 4.6 outperforms OpenAI’s GPT-5.2 by round 144 Elo factors and Claude Opus 4.5 by 190 factors.
The corporate additionally claimed that Opus 4.6 performs higher than another mannequin on BrowseComp, which it mentioned measures the power to find hard-to-find info on-line.
Anthropic mentioned its system card reveals Opus 4.6 has an total security profile pretty much as good as, or higher than, different frontier fashions, with low charges of misaligned behaviour throughout security evaluations.
GitHub chief product officer Mario Rodriguez mentioned: “Early testing reveals Claude Opus 4.6 delivering on the complicated, multi-step coding work builders face daily—particularly agentic workflows that demand planning and gear calling. This begins unlocking long-horizon duties on the frontier.”
The discharge of Claude Opus 4.6 additionally contains product and platform updates meant to help longer and extra complicated duties. In Claude Code, Anthropic mentioned customers can assemble agent groups to work on duties collectively.
On the Claude Developer Platform, the corporate launched adaptive considering, which it described as enabling the mannequin to determine when prolonged reasoning would assist, relatively than forcing builders to decide on between enabling or disabling prolonged considering.
Anthropic additionally added effort controls with 4 settings, low, medium, excessive (default) and max, that are meant to let builders steadiness intelligence, velocity and value.
To handle context limits in longer periods, Anthropic launched context compaction in beta on the API. The corporate mentioned compaction summarises and replaces older context as a dialog approaches a configurable threshold, permitting longer-running conversations and agentic duties.
Anthropic additionally mentioned Opus 4.6 helps outputs of as much as 128,000 tokens.
Outdoors developer tooling, Anthropic has made substantial upgrades to Claude in Excel and is releasing Claude in PowerPoint in a analysis preview.
The corporate mentioned the PowerPoint integration works inside the appliance by studying layouts, fonts and slide masters so edits keep aligned to current templates, and might construct slides from a company template, restructure a storyline, convert bullets into diagrams, or generate a full deck from an outline.
The PowerPoint analysis preview is in beta for Max, Workforce and Enterprise plan clients, mentioned the corporate.
Final month, stories surfaced that Anthropic was planning to boost $10bn in new funding, probably valuing the agency at $350bn. The corporate goals to considerably increase its annualised income run price this 12 months attributable to elevated adoption of its enterprise merchandise.
Join our day by day information round-up!
Give your online business an edge with our main business insights.
