Model library / RecursiveSI
Flint is a 39.8 billion-parameter language model for coding agents, using about 2.5 billion active parameters per token.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Flint is Recursive's mixture-of-experts model aimed at coding agents. It is built to explore codebases, fix bugs, implement features and write tests. The model accepts text input and output, performs reasoning and supports multi-turn tool use. It is described as suitable for agentic coding tasks and test generation.
In practice the model comprises 39.8 billion total parameters with roughly 2.5 billion active per token, a context window of 262 144 tokens and weights stored in bfloat16 format. It is released under the Apache-2.0 licence and runs on a single B300 or H200 GPU per replica using the provided SGLang serving kit. No expert layers are present in the configuration.
Specification
| Repository | RecursiveSI/flint |
| Publisher | RecursiveSI |
| Published | 21 October 2026 |
| Architecture | FlintForCausalLM |
| Model type | flint |
| Parameters | 39,785,287,456 |
| Layers | 49 |
| Hidden size | 2,048 |
| Attention heads | 32, 4 key/value heads |
| Context length | 262,144 tokens |
| Vocabulary | 150,144 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 74.1 GiB | Full precision as released |
| fp8 | 37.1 GiB | 8-bit, near-lossless on most models |
| int4 | 18.5 GiB | 4-bit, smallest footprint |
The cache costs 98 KiB per token at 16-bit, so the full 262,144-token context of one request needs about 24.5 GiB. Concurrency multiplies that number, not the weights.
That figure is an upper bound: this model uses sliding-window attention with a 513-token window on some layers, so long requests cache less than this.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 4 | 2.3 GiB | about 24,546 tokens |
| RTX 5090 32GB | 4 | 31.1 GiB | the full 262,144 tokens |
| L40S 48GB | 2 | 7.3 GiB | about 78,045 tokens |
| A100 80GB | 2 | 64.9 GiB | the full 262,144 tokens |
| H100 80GB | 2 | 64.9 GiB | the full 262,144 tokens |
| RTX PRO 6000 96GB | 1 | 9.8 GiB | about 104,794 tokens |
| H200 141GB | 1 | 50.3 GiB | the full 262,144 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve RecursiveSI/flint \ --max-model-len 20480 \ --gpu-memory-utilization 0.90
Questions
The weights take about 74 GiB at the released precision. That fits on one RTX PRO 6000 96GB, which leaves roughly 10 GiB for the key-value cache.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 262,144 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
Hy-Memory-30B-A3B is a Tencent-developed long-term-memory writer model designed to replace LLMs in Mem0, MemOS and Hy-Memory fram…
Sage-A1.5 is a 35-billion-parameter mixture-of-experts model aimed at long-horizon scientific and agentic tasks.
Xing4.0-29B-Intent is a schema-guided large language model for intent recognition and structured information extraction.
Granite Switch 4.1 30B is a modular LLM combining a 30-billion base, decision and guardian adapters, and an English speech-to-tex…
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.