Compare AI models
| Specifications | ||
ParametersThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. | 253B | — |
Context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider. | 128k | 1M |
| Benchmarks | ||
LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | 66.31% | 37.8% |
HumanEvalFunction synthesis — 164 small Python problems: the AI is given a function's description and has to write the working function. This was the coding benchmark of the GPT-3.5 and GPT-4 era, before the field moved to fixing real bugs in real repositories. Higher is better. | — | 86.6% |
IFEvalAgentic tasks — following verifiable instructions in generated responses | — | 76.3% |
MATHReasoning — accuracy on the original competition mathematics test set | — | 81.1% |
MATH-500Reasoning — accuracy on the 500-problem MATH subset | 97% | — |
AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools | 80.8% | — |
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools | 72.5% | — |
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 76.01% | — |
MMLU-ProKnowledge — accuracy on the ten-option MMLU-Pro benchmark | — | 64.5% |
GSM8KGrade-school math — Grade-school maths word problems that take a few steps of arithmetic to work through. It separated the models of 2022 and 2023 sharply, then saturated. One caveat on the historical numbers: OpenAI included part of the GSM8K training set in GPT-4's pre-training mix, so GPT-4's score is not a clean few-shot result. Higher is better. | — | 93.8% |
| Overview | ||
| Company | NVIDIA | Qwen |
| Release date | Apr 7 2025 | Nov 15 2024 |
| Access | Open Weight | Closed |
| Model details | View model | View model |
Other comparisons
Llama Nemotron Ultra 253BvsClaude Fable 5.2Qwen2.5-Turbo (1M)vsClaude Fable 5.2Llama Nemotron Ultra 253BvsGPT-6.2 AstraQwen2.5-Turbo (1M)vsGPT-6.2 AstraLlama Nemotron Ultra 253BvsGemini 4 NeonQwen2.5-Turbo (1M)vsGemini 4 NeonLlama Nemotron Ultra 253BvsMuse EmberQwen2.5-Turbo (1M)vsMuse EmberLlama Nemotron Ultra 253BvsGrok 4.7Qwen2.5-Turbo (1M)vsGrok 4.7Llama Nemotron Ultra 253BvsDeepSeek-V4.1-Flash-1017Qwen2.5-Turbo (1M)vsDeepSeek-V4.1-Flash-1017Frequently asked questions
Llama Nemotron Ultra 253B leads Qwen2.5-Turbo (1M) on 1 of the 1 benchmark they both report (LiveCodeBench). Qwen2.5-Turbo (1M) shipped 143 days before Llama Nemotron Ultra 253B, so benchmark comparisons should account for the intervening progress.
Context windows are 128k (Llama Nemotron Ultra 253B) vs 1M (Qwen2.5-Turbo (1M)). Llama Nemotron Ultra 253B is open weight, while Qwen2.5-Turbo (1M) is closed.
On LiveCodeBench, Llama Nemotron Ultra 253B leads at 66.31% vs Qwen2.5-Turbo (1M) at 37.8%.