Recursive: Flint (free)

Recursive: Flint (free) is a text model from Recursive, published under the model id recursive/flint-v1.0:free. It is an open-weight mixture-of-experts model with roughly 2.5B active parameters out of 40B total, and it was built by Recursive's autoresearch system, Eureka Machine, specifically for coding agents. The provider release date is 21 October 2026.

Flint takes text as input and returns text as output, so it fits standard chat and completion style workloads rather than image or audio tasks. Its context window is 262,144 tokens, which allows very large prompts such as long source files, multi-file repositories or extended conversation histories to be handled in a single request. The maximum output is 235,929 tokens, so long generations are possible within that window. Because it is a mixture-of-experts design, only a fraction of the total parameters are active per token, which is relevant if you care about throughput relative to a dense model of similar total size.

The combination of a large context window and a focus on coding agents points to practical uses such as code generation, refactoring and review across large files, working through repository-level context, and agent loops where the model calls tools and reasons over the results. The long maximum output also suits tasks that produce substantial text in one pass, for example generating a full module or a detailed explanation of a codebase. Outside of coding, the same text-in, text-out setup works for general assistant and analysis tasks, provided the input is text.

You call the model through api14's OpenAI-compatible API using the model id recursive/flint-v1.0:free. The supported request parameters are include_reasoning, logprobs, max_tokens, reasoning, response_format, seed, stop, temperature, tools, top_logprobs and top_p. The presence of tools and response_format makes structured output and tool calling straightforward, while seed, temperature, top_p, stop and max_tokens cover the usual sampling and length controls. The reasoning and include_reasoning parameters relate to the model's reasoning behaviour, and logprobs with top_logprobs let you inspect token-level probabilities.

The model is text only, so if you need image or audio input or output, a different model is required. It is also offered as a free variant, which may mean availability or throughput differs from paid options, though no specific limits are stated here. If your workload is small and does not benefit from the large context window or coding-agent focus, a simpler text model may be a better fit.