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. | 9B/27B | 8B |
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. | 8k | 8k |
| API pricingUSD per 1M tokens · lower wins | ||
Cheapest inputLowest input rate across every provider on OpenRouter, the lab included. The cheapest endpoint may run a quantised build or a shorter context — see "Available from" on the model page. | $0.65NextBit | — |
Cheapest outputLowest output rate across every provider on OpenRouter, the lab included. May come from a different provider than the cheapest input. | $0.65NextBit | — |
| $163NextBit · int4 | — | |
| Benchmarks | ||
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. | 51.8% | 36.22% |
ARC-ChallengeReasoning — multiple-choice science questions from the AI2 Reasoning Challenge; unrelated to ARC-AGI | 71.4% | 64.42% |
HellaSwagReasoning — commonsense completion of short scenarios | 86.4% | 83.03% |
MMLUGeneral knowledge — A 57-subject multiple-choice exam — history, law, medicine, maths — that was the standard measure of how much a model knows from 2020 until roughly 2024, when frontier scores crowded into the high 80s and labs moved on to harder tests. The scores here were published years apart under different testing setups, so read them as a historical record rather than a like-for-like ranking. Higher is better. | 75.2% | 69.51% |
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. | 74% | 58.45% |
MATHReasoning — accuracy on the original competition mathematics test set | 42.3% | — |
WinoGrandeReasoning — resolving ambiguous pronouns in context | — | 80.35% |
| Overview | ||
| Company | NVIDIA | |
| Release date | Jun 27 2024 | Aug 21 2024 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
Gemma 2vsClaude Fable 5.2Mistral NeMo Minitron 8BvsClaude Fable 5.2Gemma 2vsGPT-6.2 AstraMistral NeMo Minitron 8BvsGPT-6.2 AstraGemma 2vsMuse EmberMistral NeMo Minitron 8BvsMuse EmberGemma 2vsGrok 4.7Mistral NeMo Minitron 8BvsGrok 4.7Gemma 2vsDeepSeek-V4.1-Flash-1017Mistral NeMo Minitron 8BvsDeepSeek-V4.1-Flash-1017Gemma 2vsMistral Large 4Mistral NeMo Minitron 8BvsMistral Large 4Frequently asked questions
Gemma 2 leads Mistral NeMo Minitron 8B on 5 of the 5 benchmarks they both report. Gemma 2 shipped 55 days before Mistral NeMo Minitron 8B, so benchmark comparisons should account for the intervening progress.
Gemma 2 has 9B/27B parameters, while Mistral NeMo Minitron 8B has 8B. Context windows are 8k (Gemma 2) vs 8k (Mistral NeMo Minitron 8B).
On HumanEval, Gemma 2 leads at 51.8% vs Mistral NeMo Minitron 8B at 36.22%. On ARC-Challenge, Gemma 2 leads at 71.4% vs Mistral NeMo Minitron 8B at 64.42%. On HellaSwag, Gemma 2 leads at 86.4% vs Mistral NeMo Minitron 8B at 83.03%. On MMLU, Gemma 2 leads at 75.2% vs Mistral NeMo Minitron 8B at 69.51%. On GSM8K, Gemma 2 leads at 74% vs Mistral NeMo Minitron 8B at 58.45%.