dcostenco/prism-coder-14b

TEXT GENERATIONConcurrency Cost:1Model Size:14.8BQuant:FP8Ctx Length:32kTool Calling:SupportedPublished:May 4, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

dcostenco/prism-coder-14b is a deprecated language model based on Qwen 3 architecture, previously intended for coding tasks. It has been superseded by dcostenco/prism-coder-9b, which offers improved performance and a smaller model size. This 14B variant is no longer recommended for use, with the 9B replacement achieving 100% on BFCL Overall and a 36% reduction in size.

Loading preview...

Model Status: Deprecated

dcostenco/prism-coder-14b is a deprecated model. It has been replaced by prism-coder-9b, which is based on the Qwen 3.5 architecture and offers significant improvements.

Key Differences from Replacement (prism-coder-9b)

  • Performance: The 14B model achieved 90.3% on BFCL Overall, while the 9B replacement achieves 100.0%.
  • Model Size: The 14B model is 9.0 GB, whereas the 9B replacement is 5.8 GB, representing a 36% reduction in size.
  • Hallucinations: Both models maintain 0 reported hallucinations.

Recommended Alternatives

Users are strongly advised to use dcostenco/prism-coder:9b for general coding tasks, or explore other models in the dcostenco/prism-coder fleet for specific use cases:

  • dcostenco/prism-coder:2b (Qwen3.5-2B): Optimized for mobile devices (iPhone).
  • dcostenco/prism-coder:4b (Qwen3.5-4B): Suitable as a verifier model for systems with 8 GB+ RAM.
  • dcostenco/prism-coder:9b (Qwen3.5-9B): The default recommended router for most applications.
  • dcostenco/prism-coder:32b: Designed for more complex tasks requiring higher capacity.