Anthropic ships Opus 4.8, with Mythos following soon

@AnthropicAI has released Claude Opus 4.8, an upgrade to its flagship model that arrives at no extra cost to existing users. Standard pricing remains at $5 per million input tokens and $25 per million output tokens, identical to Opus 4.7.

Stronger Benchmarks, More Honest Outputs

Opus 4.8 outperforms its predecessor across most major benchmarks and beats OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro in several key categories. On agentic coding (SWE-Bench Pro), it scores 69.2%, compared to 64.3% for Opus 4.7, 58.6% for GPT-5.5, and 54.2% for Gemini 3.1 Pro. In GDPval, a benchmark that judges an AI agent's ability to complete economically viable work, Opus 4.8 scored 1890, a jump from Opus 4.7's 1753 and GPT-5.5's 1769.

Beyond raw scores, the behavioral shift may matter more to enterprise users. Early testers found the new model is more likely to flag uncertainties about its work and less likely to make unsupported claims. A Bridgewater testimonial highlighted the model's tendency to proactively flag issues with the inputs and outputs of an analysis, something other models left to users to catch.

New features ship alongside the model. Anthropic is launching a dynamic workflow tool that lets Claude run multiple subagents at once, and a control panel that lets users decide how much effort Claude puts into a response. Fast Mode runs at approximately 2.5x the speed and is priced at one-third the previous cost.

The Crypto Angle: Mythos is Coming

The new model still lags the performance of Mythos, Anthropic's most advanced, but the company says Mythos-class models are expected in the coming weeks. That timeline carries weight for the crypto industry. Anthropic announced the limited release of Claude Mythos Preview on April 7, 2026, restricting access through Project Glasswing, a controlled program offering the model only to select tech firms and key infrastructure providers for defensive cybersecurity purposes.

Anthropic's Claude Mythos Preview has autonomously discovered serious zero-day vulnerabilities in widely used software, outperforming both human researchers and existing automated tools. For years, DeFi has focused its defenses on smart contracts, with code audited and common exploits well understood. But Mythos, designed to identify and chain together weaknesses across systems, is pushing attention beyond code and into the infrastructure that supports it. Coinbase and Binance have both reportedly approached Anthropic to test Mythos.

With broad access now confirmed for the coming weeks, the Opus 4.8 release itself comes just 41 days after Opus 4.7, signaling a faster iteration pace at Anthropic as competition from OpenAI and Google intensifies.

Sources:
TechCrunch: Anthropic releases Opus 4.8 with new dynamic workflow tool
CoinDesk: How Anthropic's Mythos model is forcing the crypto industry to rethink everything about security
Axios: Anthropic releases new model, Opus 4.8