Claude Sonnet 5 Launches at $2 per Million Tokens, Nearing Opus Performance
Anthropic released Claude Sonnet 5: near-Opus 4.8 performance at $2/$10 per million tokens until Aug 31, a native 1M-token context window, and default status in Claude Code.
Anthropic released Claude Sonnet 5: near-Opus 4.8 performance at $2/$10 per million tokens until Aug 31, a native 1M-token context window, and default status in Claude Code.
Anthropic released Claude Sonnet 5 on June 30. It moves closer to Opus 4.8 on benchmarks while keeping the price flat or lower, and it’s already the default for Free and Pro users.
Claude Code v2.1.181 adds /config key=value so you can flip settings straight from the prompt — in interactive, pipe, and remote modes.
Claude Code’s Week 24 update adds fallbackModel — when your main model is overloaded, it automatically switches to a backup, with up to three models tried in order.
Claude Code’s new –safe-mode flag (v2.1.169) boots with all your customizations off, so you can isolate which setting broke things.
Warp is a Rust-built terminal that groups each command and its output into a block, with a free tier, AI credits, and a growing agentic mode. A breakdown from the official docs.
Starting June 15, programmatic Claude usage (Agent SDK, claude -p, GitHub Actions) draws from separate Agent SDK credits, not your Pro/Max subscription. Interactive use is unchanged.
Claude Code 2.1.174 adds a usage attribution view that breaks down where your tokens go: cache misses, subagents, skills, plugins, and MCP. VS Code only for now.
Claude Code 2.1.172 lets subagents spawn their own subagents, nesting up to 5 levels deep. Here’s what changes for task delegation — and the token cost to watch out for.
Claude Code 2.1.152 added /reload-skills, which rescans your skill directory without killing the session. For anyone editing skill files daily, it ends the restart-and-rebuild-context loop.