AI Models
Flint
Open-weight mixture-of-experts model for high-speed local agentic coding
Open sourceRuns in the cloudEditors pick
Pricing
Open Source
Free
Platforms
Web
Hosted app
Category
AI Models
22 tools
Reviews
0
No verified reviews yet
Last updated
Oct 24
Listing updated
Best for
High-speed local agentic coding and test-driven debugging
Execute terminal commands, run test suites, and iterate on code fixes locally at over 120 tokens per second.
Pricing checked Oct 2026
User feedback
No aggregate rating verified
Explore discussions, issues, and documentation to see how this project fits your workflow.
Prother ratings come from verified reviews only.
Read Prother reviews (0) - write your ownHide Prother reviews
Reviews (0)
HONEST AND VERIFIED ONLYReview Flint
Share real experiences on ease of use, feature power, and overall value.
About Flint
Flint is an open-weight Mixture-of-Experts (MoE) foundation model created by Recursive Superintelligence. Spanning 40 billion total parameters while activating only 2.5 billion per token, Flint is specifically architected to deliver frontier-level software engineering performance on local workstations and laptops.
Trained with Recursive's automated research systems and tuned across software benchmark harnesses, Flint can exceed 125 tokens per second on consumer Apple Silicon and RTX hardware. It provides deep terminal tool-use proficiency, rapid repository navigation, and autonomous multi-turn bug resolution under an Apache 2.0 license.
- Sparse 2.5B active parameter footprint delivers blazing local inference speed
- Permissive Apache 2.0 open-source license with weights on Hugging Face
- Tuned specifically for terminal navigation, tool execution, and code synthesis
- MoE memory layout still requires ~24GB VRAM or unified memory for uncompressed weights
- Focused heavily on coding tasks rather than general prose or creative writing
Practical Guides & In-Depth Evaluations
More in AI Models
See categoryFrontier foundation models, large language models, open-weight reasoning architectures, and multimodal intelligence systems. Every listing evaluates benchmark performance, native context window limits, parameter scale, licensing terms, and deployment efficiency across local hardware and managed APIs.
