KAT Dev 32B
accounts/fireworks/models/kat-dev-32b
Ready
Tunable
LLM
KAT-Dev-32B is an open-source 32B-parameter model for software engineering tasks. It is optimized via several stages of training, including a mid-training stage, supervised fine-tuning (SFT) & reinforcement fine-tuning (RFT) stage and an large-scale agentic reinforcement learning (RL) stage.
Details
StateReady
Created onNov 15, 2025
KindBase model
ProviderKwaipilot
Features
CalibratedNo
Mixture-of-ExpertsNo
Parameters32B
Fine-tuningSupported (LoRA)
ServerlessNot supported
Context Length131k tokens
Function CallingNot supported
EmbeddingsNot supported
RerankersNot supported
Image InputNot supported
Usage guide
1. Choose how you’d like to use KAT Dev 32B
2. Create a deployment
Deploy KAT Dev 32B in just a few commands.
- No rate limits
- Lower cost-of-ownership for high-traffic
- Large selection of models and custom base models
Dedicated deployments on Fireworks
On-demand deployments ensure you get dedicated GPUs for optimal performance, no rate limits, and fast autoscaling. Fireworks makes it easy to get started in just minutes.
Dedicated deployments are great for high utilization use cases – unlocking both cost-effective scale and better performance.
- Get started in minutes with our Deployment Quickstart guide
- Learn how to customize your autoscaling configuration
- Check out our advanced features (e.g. speculative decoding, quantizations, performance benchmarking, and more)