Anthropic has officially launched Claude Opus 4.8 – a direct challenger to OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro. The release comes with a primary focus on agentic coding, large-scale automation and improved reasoning. The company also says the new model is significantly ‘more honest and less likely to make unsupported claims during complex workflows’.
According to Athropic, Opus 4.8 is available globally immediately and keeps the same pricing structure as Opus 4.7 despite broad capability upgrades.
Claude Opus 4.8 vs GPT 5.5 vs Gemini 3.1 Pro
Anthropic published a direct comparison chart in its press release. While Claude Opus 4.8 leads most benchmark categories overall, GPT-5.5 actually outperformed it in one critical area: agentic terminal coding.
The benchmark image shows:
Claude Opus 4.8 scored 69.2% on SWE-Bench Pro for agentic coding, ahead of GPT-5.5’s 58.6% and Gemini 3.1 Pro’s 54.2%.
GPT-5.5 led Terminal-Bench 2.1 agentic terminal coding with 78.2%, beating Opus 4.8’s 74.6%.
Opus 4.8 posted the highest multidisciplinary reasoning scores both with and without tools.
In OSWorld-Verified agentic computer use, Opus 4.8 narrowly led with 83.4%.
For knowledge work tasks measured by GDPval-AA, Opus 4.8 achieved 1890 compared to GPT-5.5’s 1769 and Gemini’s 1314.
Opus 4.8 also topped agentic financial analysis benchmarks at 53.9%.
The image also highlights how close the top models now are in several categories, especially between Claude Opus 4.8 and GPT-5.5.
Anthropic emphasizes ‘honesty improvements’
One of Anthropic’s biggest talking points around the release is reliability. The company said early testers found Opus 4.8 ‘more reliable and sharper in its judgement’ during agentic tasks. Anthropic specifically focused on reducing hallucinations.
The company stated that Opus 4.8 is ‘around four times less likely’ than Opus 4.7 to allow flawed code to pass without flagging issues. Anthropic also said the model now ‘reaches new highs on our measures of prosocial traits like supporting user autonomy and acting in the user’s best interest’.
The company added that misaligned behaviors such as deception or cooperation with misuse are ‘substantially lower’ than Opus 4.7.
New effort control changes how users interact with Claude
A major addition alongside the release is customizable ‘effort’ settings. Users on Claude.ai can now determine how much computational power the model spends on a task. Lower effort settings generate faster answers while consuming fewer tokens. Higher settings force the model to reason more deeply before responding.
Anthropic said: “Users now have this choice, the effort control is available on all plans.”
The company noted that Opus 4.8 defaults to ‘high effort’, which it considers the best balance between speed and output quality. For difficult workflows, users can select ‘extra’ or ‘max’ effort modes that allocate more tokens toward deeper reasoning.
Dynamic workflows push AI agents further
Anthropic simultaneously introduced a research-preview feature called ‘dynamic workflows’ inside Claude Code. The feature allows the system to coordinate hundreds of parallel AI subagents within one coding session.
Anthropic described the capability by saying: “Claude can plan the work and then run hundreds of parallel subagents in a single session.”
According to the company, the feature is capable of handling ‘codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge’.
Claude Opus 4.8 price
Despite the performance increases, Anthropic did not raise standard pricing.
The company confirmed pricing remains:
$5 per million input tokens
$25 per million output tokens
Fast mode pricing is listed at:
$10 per million input tokens
$50 per million output tokens
Anthropic also said fast mode for Opus 4.8 is now ‘three times cheaper’ operationally than previous versions while delivering responses at 2.5× the speed.
