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. | 7B/13B/70B | 6B |
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. | 4k | 8k |
| 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. | 29.9% | 58.5% |
MATHReasoning — accuracy on the original competition mathematics test set | 13.5% | 25.7% |
HellaSwagReasoning — commonsense completion of short scenarios | 85.3% | 79.7% |
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. | 68.9% | 61.4% |
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. | 56.8% | 72.3% |
ARC-ChallengeReasoning — multiple-choice science questions from the AI2 Reasoning Challenge; unrelated to ARC-AGI | 57.4% | — |
WinoGrandeReasoning — resolving ambiguous pronouns in context | 80.2% | — |
| Overview | ||
| Company | Meta | Z.ai |
| Release date | Jul 18 2023 | Oct 27 2023 |
| Access | Open Weight | Open Weight |
| Model details | View model | View model |
Other comparisons
LLaMA 2vsClaude Fable 5.2ChatGLM3-6BvsClaude Fable 5.2LLaMA 2vsGPT-6.2 AstraChatGLM3-6BvsGPT-6.2 AstraLLaMA 2vsGemini 4 HydrogenChatGLM3-6BvsGemini 4 HydrogenLLaMA 2vsGrok 4.7ChatGLM3-6BvsGrok 4.7LLaMA 2vsDeepSeek-V4.1-Flash-1017ChatGLM3-6BvsDeepSeek-V4.1-Flash-1017LLaMA 2vsMistral Large 4ChatGLM3-6BvsMistral Large 4Frequently asked questions
ChatGLM3-6B leads LLaMA 2 on 3 of the 5 benchmarks they both report. LLaMA 2 shipped 101 days before ChatGLM3-6B, so benchmark comparisons should account for the intervening progress.
LLaMA 2 has 7B/13B/70B parameters, while ChatGLM3-6B has 6B. Context windows are 4k (LLaMA 2) vs 8k (ChatGLM3-6B).
On HumanEval, ChatGLM3-6B leads at 58.5% vs LLaMA 2 at 29.9%. On MATH, ChatGLM3-6B leads at 25.7% vs LLaMA 2 at 13.5%. On HellaSwag, LLaMA 2 leads at 85.3% vs ChatGLM3-6B at 79.7%. On MMLU, LLaMA 2 leads at 68.9% vs ChatGLM3-6B at 61.4%. On GSM8K, ChatGLM3-6B leads at 72.3% vs LLaMA 2 at 56.8%.