Flint 40B

flint-40b 🇮🇹 Italy

Recursive's model for coding agents (October 2026): it looks through a project, fixes bugs and writes tests, calling your functions as it goes. Only 2.5 of its 40 billion parameters work on each token, so it is quick. It thinks before answering, without showing it: the thinking counts in the tokens. Context of 65,536 tokens, 4-bit NVFP4 weights from the Ollama library. Runs on NVIDIA DGX Spark (GB10), in Italy. In service.

At a glance

Publisher	Recursive
Architecture	Mixture of experts, 40 billion parameters, about 2.5 billion per token
Precision	NVFP4, 4 bits (Ollama library)
Context	65,536 tokens, prompt and answer together
Longest answer	`max_tokens` up to 32,768, reasoning included
Reads	Text
Calls tools	Yes
JSON schema	Yes
Reasons	Yes, always, without showing it: the reasoning counts as output tokens.
Runs in	🇮🇹 Italy
Status	Loaded: the first request answers without a start-up wait. Live state on Status.

Use it

curl https://api.ai.itsincom.org/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "flint-40b",
    "messages": [{"role": "user", "content": "Summarise in one line: the meeting moves to Thursday at 10."}]
  }'

import os
from openai import OpenAI

client = OpenAI(base_url="https://api.ai.itsincom.org/v1", api_key=os.environ["API_KEY"])
answer = client.chat.completions.create(
    model="flint-40b",
    messages=[{"role": "user", "content": "Summarise in one line: the meeting moves to Thursday at 10."}],
)
print(answer.choices[0].message.content)

More in Quickstart and Chat completions.

Limits

Prompt and answer together stay within 65,536 tokens; beyond that the engine refuses the request with a 400 that says so.
The reasoning counts against max_tokens: a low cap can end the answer before it starts, with finish_reason: "length".
Text only: a request with an image gets a 400.
Served at 4 bits, which makes it faster and fits our machines; we write the precision of every model.
A request goes only to the zones your key allows; if the model has no machine there now, a 503 names the zones. It is never sent elsewhere.