Anthropic releases Claude Opus 4.8 with improved agentic coding and reasoning benchmarks
Anthropic released Claude Opus 4.8 on May 28, 2026, with improved performance in agentic coding, computer use, and reasoning benchmarks. Pricing remains at $5 per million input tokens and $25 per million output tokens, while the model's fast mode is now three times cheaper than previous versions.
Claude Opus 4.8 — Quick Specs
Claude Opus 4.8 Released with Agentic Improvements
Anthropic released Claude Opus 4.8 on May 28, 2026, positioning it as a "modest but tangible improvement" over Opus 4.7 with a focus on agentic capabilities.
Pricing and Performance
The model maintains the same pricing as Opus 4.7: $5 per million input tokens and $25 per million output tokens. According to Anthropic, fast mode for Opus 4.8—which operates at 2.5× standard speed—is now three times cheaper than in previous models.
Anthropic's internal benchmarks show improvements over Opus 4.7 in three key areas:
- Agentic coding
- Computer use
- Reasoning tasks
Specific benchmark scores were not disclosed in the announcement.
New Features
The release introduces several capabilities:
Effort control: Users on claude.ai can now adjust how much computational effort Claude applies to a given task.
Dynamic workflows: Claude Code, Anthropic's coding-focused interface, gains a new feature designed to handle large-scale problems by breaking them into manageable workflows.
Enhanced usage limits: Claude Code users receive doubled usage limits, though specific numbers were not provided.
Context and Competition
The release comes one week after Google I/O, where Google announced Gemini 3.5 Pro with a similar focus on agentic capabilities. Anthropic did not provide direct comparisons between Opus 4.8 and Gemini 3.5 Flash, noting they represent different model classes. Some comparisons appear in Anthropic's Opus 4.8 system card, though details were not included in the announcement.
What This Means
This incremental release signals Anthropic's continued focus on practical AI applications, particularly autonomous task execution. The unchanged pricing suggests Anthropic views the improvements as evolutionary rather than requiring a new pricing tier. The emphasis on agent capabilities aligns with broader industry trends, as both Anthropic and Google position autonomous AI systems as a key competitive battleground for 2026. The three-fold price reduction for fast mode could make high-speed inference more accessible for production deployments.
Related Articles
Anthropic Paper: Automated AI Researchers Beat Humans at Alignment Fixes for $4/Hour
A new Anthropic paper from its fellows program shows an automated AI system improving performance on all 10 tested alignment benchmarks, outperforming experienced human researchers within six hours at a fraction of the cost. The research, led by Anthropic Fellow Chen Yueh-Han, is described as early evidence that automated alignment post-training could become practical soon.
Anthropic Adds Built-In Browser to Claude Cowork Desktop App
Anthropic is embedding a dedicated browser into Claude Cowork's desktop app, opening in a side panel whenever a task requires web access. The browser is isolated from the user's own tabs, bookmarks, and passwords, and rolls out this week to Pro, Max, Team, and Enterprise plans.
Alibaba Releases Qwen3.8-Flash-Next, a 125B-Parameter Preview of Qwen4's Architecture
Alibaba's Qwen team has released Qwen3.8-Flash-Next, an open-weight model with 125B total parameters (6B activated) that previews architectural changes planned for Qwen4, including a new sparse attention mechanism and n-gram embeddings. The model natively supports 262,144 tokens of context, extensible to 1 million.
IBM Releases Granite 4.2 Open-Weight Models With Agentic RL Training and 512K Context
IBM has released Granite 4.2, a family of open-weight language models in 3B, 8B, and 30B parameter sizes, trained on roughly 15 trillion tokens with context windows up to 512,000 tokens. The 8B and 30B variants underwent additional 'agentic RL' training for tool use, code execution, and web search.
Comments
Loading...