Claude Opus 5 support lands in T3 Code
T3 Code added Claude Opus 5 model support in a 24 July 2026 nightly, days after the model shipped — a good illustration of why the CLI-wrapper architecture pays off.
24 July 2026 — build v0.0.29-nightly.20260724.896 added Claude Opus 5 model support.
A single changelog line, but it says something useful about how the project is built.
What changed#
Claude Opus 5 is now selectable in the composer's model picker when you're using the Claude Code provider. The same batch of builds restored the model picker layout and added a bolt icon for fast mode, so the whole model-selection surface got attention in the same week.
To use it:
# make sure your Claude Code CLI is current
npm i -g @anthropic-ai/claude-code@latest
claude auth login
# and T3 Code
npx t3@latestThen pick Opus 5 in the composer. If it isn't listed, one of the two is out of date — that's the usual cause across every model-availability question.
Why it arrived so fast#
This is the architectural argument in miniature. T3 Code doesn't integrate models — it speaks JSON-RPC to provider CLIs, and the CLI handles model access.
T3 Code → Claude Code CLI → Anthropic
(UI, git) (agent, models)So supporting a new model is mostly surfacing something the CLI already exposes, not building an integration. That's why a wrapper can track a new frontier model within days, and why four providers were supported so quickly.
The trade-off is the same coupling that causes the version-skew failures T3 Code is known for. You inherit the upside and the downside of being downstream.
Switching models inside a thread#
The feature this makes more useful: you can change models within a workspace thread rather than starting over. The practical pattern —
- Exploration and boilerplate — a fast, cheap model. Reading a codebase, scaffolding, mechanical edits.
- The part that matters — switch to Opus 5 for the design decision, the tricky refactor, the thing you'd hate to get subtly wrong.
You keep the accumulated context either way, which is the whole point. Doing this across separate terminal sessions means re-establishing context each time.
Worth watching your usage during the first week, though. Agentic runs read far more than chat conversations do, and a capable model across many turns adds up faster than people expect.
Checking what your build supports#
Model availability depends on three things, and it's worth knowing which to check:
- Your provider CLI version — the CLI must know about the model
- Your T3 Code version — the picker must surface it
- Your provider account — the model must be available on your plan
When a model doesn't appear, work through those in order. Two of the three are fixed by updating.
FAQ#
Does T3 Code support Claude Opus 5?#
Yes. Claude Opus 5 model support was added in the 24 July 2026 nightly build v0.0.29-nightly.20260724.896, selectable in the composer's model picker with the Claude Code provider.
How do I use Claude Opus 5 in T3 Code?#
Update both the Claude Code CLI and T3 Code to current versions, authenticate with claude auth login, then select Opus 5 in the composer's model picker.
Why doesn't the new model appear in my model picker?#
Usually an outdated Claude Code CLI or an outdated T3 Code build. Update both. It can also mean the model isn't available on your provider plan.
Can I switch models mid-thread in T3 Code?#
Yes. You can change models and agents within a workspace thread while keeping the context you've already built up.