Skip to model comparison
AI Agent Store/Model comparisonsBuild an AI worker ↗

Recursive · Download verified

Recursive Flint · download

Run a coding model on your own hardware to inspect a codebase, make changes and write tests.

First reviewed 2026-10-23 · Checked 2026-11-03. The launch date has not been confirmed.

We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Try it on your hardware and tasks before relying on it.

Runs onYour hardwareHardware and setup costs apply
Documented capacity262KSee input limits below

What you can use today

You download the model files and supply the hardware and software to run them. Capacity is the amount of information a request can contain; settings and available memory can limit what fits.

Exact download ID: RecursiveSI/flint. Text input and output, reasoning and tool calls. Preserve the returned reasoning history between turns. A verified adjustable effort menu is not available.

Apache 2.0 weights. The BF16 download is about 80 GB before runtime memory; the publisher supplies an SGLang kit for B300/H200 GPUs. Laptop use needs a suitable compressed format and engine.

Prompt, thinking and answer share 262,144 tokens. The suggested 32,768-token output budget is an example, not a universal maximum.

No local per-token price: include hardware and operating costs. Publisher benchmark and speed claims are not independent results.

Provider release and specifications ↗

Check availability and plans ↗

Provider documentation ↗

Test it before switching

Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Include the time and hardware needed to run it.

Compare caching and long-context costs ↗
AI Agent Store · Model comparisons · Data reviewed 2026-11-03Sources, limitations & corrections ↗