Claude Opus 4.8 is live in Windsurf. Standard pricing stays at $5.00/million input tokens and $25.00/million output tokens — identical to Opus 4.7. The real move is Fast Mode.
Here’s what changed with Opus 4.8:
- Standard pricing unchanged — $5.00 input / $25.00 output per million tokens, same as Opus 4.7
- Fast Mode price dropped ~3x — previous Opus fast modes ran roughly $30/$150 per million tokens; Opus 4.8 Fast Mode lands at $10.00/$50.00
- 1M token context window — available by default on the Claude API, Amazon Bedrock, and Vertex AI (200k on Microsoft Foundry)
- 128k max output tokens — with adaptive thinking and the same tool support as Opus 4.7
The Fast Mode price cut is the part worth paying attention to. At $10/$50 per million tokens, it’s a third of what Anthropic charged for comparable speed tiers on earlier Opus versions. For Windsurf users running agentic coding sessions — where the model churns through large codebases and generates long outputs — that gap compounds fast.
Opus 4.8 is also available across Cursor, JetBrains IDEs, Kilo CLI, and Trae, so Windsurf isn’t the only IDE picking it up. But Windsurf’s AI-first workflow leans harder on extended model interactions, which makes the Fast Mode economics more relevant here than in editors where you’re mostly doing tab-completions.
NOTE
Anthropic’s official Opus 4.8 page lists adaptive thinking and the effort parameter as carried over from 4.7. No new tool capabilities were added — this is a speed and cost update, not a capability jump.
Fast Mode at $10/$50 per million tokens makes extended agentic sessions in Windsurf roughly 3x cheaper than they were on previous Opus fast tiers. Available now — no waitlist, no feature flag.