Bolt: Claude Opus 4.6 with Variable Effort Reasoning
BoltBolt integrated Claude Opus 4.6 into its platform following Anthropic's February 5, 2026 release, adding a variable effort selection feature that lets developers choose between lighter and deeper reasoning modes depending on task complexity. Bolt's agent paired with Opus 4.6 can break work into parallel streams, reducing the number of prompts needed to complete larger builds. The model's 1 million token context window makes it especially well-suited for projects involving large codebases.
Sources & Mentions
5 external resources covering this update
Anthropic releases Opus 4.6 with new 'agent teams'
TechCrunch
Anthropic's Claude Opus 4.6 brings 1M token context and 'agent teams' to take on OpenAI's Codex
VentureBeat
Harvey, Bolt, Shopify & Lovable Test Claude Opus 4.6 Pre-Launch
Anthropic
Claude Opus 4.6 (discussion)
Hacker News
Anthropic Releases Claude Opus 4.6 With 1M Context, Agentic Coding, Adaptive Reasoning Controls
MarkTechPost
Claude Opus 4.6 Arrives in Bolt
On February 5, 2026, Anthropic released Claude Opus 4.6 — and Bolt made it available to users shortly after. The integration brings Anthropic's most capable model directly into the Bolt build environment, where it powers the platform's AI agent for coding, debugging, and full-stack application development.
Garrett Serviss, VP of Marketing at Bolt, noted that Opus 4.6 "diagnosed bugs on the first try that we'd failed to fix across five-plus attempts with previous models," pointing to a measurable improvement in reasoning depth for complex debugging scenarios.
Variable Effort: Matching Reasoning Depth to Task Complexity
The most developer-relevant addition in the Bolt integration is variable effort selection — a feature tied to Opus 4.6's adaptive reasoning system. Rather than applying the same level of reasoning to every task, Bolt users can now choose how hard the model "thinks" based on what they're building.
Lighter effort settings are faster and consume fewer tokens — ideal for quick UI tweaks, copy edits, or straightforward component generation. Deeper effort settings activate more thorough reasoning, useful for debugging complex logic, planning multi-file refactors, or generating backend architecture decisions. This dynamic allocation makes the AI more cost-efficient for routine tasks while preserving its full reasoning capability for hard problems.
Expanded Context Window
Opus 4.6 ships with a 1 million token context window, a significant jump from the 200,000 token limit on previous Opus versions. For Bolt users, this means the model can hold more of a project's code, conversation history, and imported context in a single session — reducing the context loss that was a friction point in longer build workflows.
Agent Teams Architecture
Beyond the base model improvements, Opus 4.6 introduces an "agent teams" capability where multiple agents split larger tasks into parallel workstreams and coordinate directly with each other. In Bolt's implementation, this means the AI agent can break a complex build request into concurrent sub-tasks, execute them simultaneously, and synthesize the results — delivering larger and more complete implementations with fewer round-trips from the developer.