Code reasoning focus
According to the model card, this is a reasoning model post-trained for code generation with a reported 64k-token context length.
Open Source Model Profile · nvidia
OpenCodeReasoning-Nemotron-1.1-14B
OpenCodeReasoning-Nemotron-1.1-14B is a 14.77B-parameter Qwen2 text-generation model from nvidia. According to the model card, it is a Qwen2.5-14B-Instruct derivative post-trained for reasoning for code generation.
- Publisher
- nvidia
- Task
- text-generation
- Model type
- qwen2
- License
- Unknown
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
OpenCodeReasoning-Nemotron-1.1-14B is published by nvidia as a Transformers text-generation model. The captured configuration identifies Qwen2ForCausalLM with model type qwen2 and about 14.77B Safetensors parameters. According to the model card, it is a derivative of Qwen2.5-14B-Instruct post-trained for reasoning for code generation.
Recorded capabilities
Qwen2 14B-class build
Captured configuration identifies Qwen2ForCausalLM with 14770033664 Safetensors parameters, and the model card describes a dense decoder-only Transformer with 14B parameters.
Publisher-reported LiveCodeBench table
According to the model card, the reported LiveCodeBench v5 average over 64 evaluations lists this model at 65.9 Pass@1 alongside 7B, 14B, and 32B peers.
Documented inference pattern
The model card documents Transformers and torch inference with a think-step-by-step coding prompt passed as user-role messages.
OpenCodeReasoning training corpus
According to the model card, the training corpus is the OpenCodeReasoning dataset of competitive programming questions with DeepSeek-R1-0528 responses, described as 1.165M samples.
Use cases in the source record
- Code-generation experiments using the model card's documented step-by-step reasoning prompt and Transformers inference pattern.
- Competitive-coding research comparisons that reproduce or extend the publisher's LiveCodeBench v5 evaluation setup.
Limitations and unknowns
- Benchmark figures are publisher-reported model-card claims; no Ethen-measured evaluation results were extracted.
- No pricing or current provider availability was extracted; provider state is historical snapshot data.
- Hardware notes are limited to the publisher's NeMo 2.3.0, Ampere and Hopper compatibility, Linux support, and vLLM testing on H100-80GB; no general VRAM requirement was extracted.
- Licensing mentions both the NVIDIA Open Model License Agreement as governing terms and Apache License Version 2.0 as additional information, so exact terms need confirmation.
Source and provenance
Source: nvidia/OpenCodeReasoning-Nemotron-1.1-14B
Captured: Unknown. Processed: 2026-09-07T19:35:59.388374+00:00.
OpenCodeReasoning-Nemotron-1.1-14B Overview Description: OpenCodeReasoning-Nemotron-1.1-14B is a large language model (LLM) which is a derivative of Qwen2.5-14B-Instruct (AKA the reference model). It is a reasoning model that is post-trained for reasoning for code generation. The model supports a context length of 64k tokens. This model is ready for commercial/non-commercial use. Results Below results are the average of 64 evaluations on LiveCodeBench (v5) [2408-2501]. Model Pass@1 DeepSeek-R1-0528 73.4 DeepSeek-R1 65.6 QwQ-32B 61.3 Distilled 7B+ Models Bespoke-Stratos-7B 14.7 OpenThinker-7B 25.5 R1-Distill-Qwen-7B 38.0 OlympicCoder…
F001F002F003F004F005F006F008F009F010F012F013F014F015F016F017F018F019F020F021F022F023