SAVRN Model Hub · Comparisons
Mistral-7B-Instruct-v0.2 vs Qwen2.5-Coder-7B-Instruct
Mistral-7B-Instruct-v0.2 has 7.2B parameters and Qwen2.5-Coder-7B-Instruct has 7.6B parameters; both are released under Apache License 2.0; at 16-bit, Mistral-7B-Instruct-v0.2 needs about 17.4 GB (1x MI300X from $1.85 an hour) and Qwen2.5-Coder-7B-Instruct about 18.3 GB (1x MI300X from $1.85 an hour).
| Field | Mistral-7B-Instruct-v0.2 mistralai/Mistral-7B-Instruct-v0.2 | Qwen2.5-Coder-7B-Instruct Qwen/Qwen2.5-Coder-7B-Instruct |
|---|---|---|
| Publisher | Mistral AI_ | Qwen |
| Task | Text generation | Text generation |
| Modality | Text | Text |
| Parameters, as reported | 7.2B parameters | 7.6B parameters |
| Architecture | MistralForCausalLM | Qwen2ForCausalLM |
| Library | transformers | transformers |
| Context length | 32,768 tokens | 32,768 tokens |
| Repository size | 29.5 GB | 15.2 GB |
| Artifact formats | safetensors, pytorch | safetensors |
| License | apache-2.0 | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 17.4 GB | 18.3 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 4.3 GB | 4.6 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 63a8b0818953 | c03e6d358207 |
| Downloads reported by the hub | 1.8M | 2.7M |
| Last observed | 2026-09-18 | 2026-09-18 |
An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.
Other Reported Results
These results are listed for each model on its own, because the conditions needed to compare them are not stated or do not match. Two results that leave a condition blank are not assumed to share it.
Mistral-7B-Instruct-v0.2
| Benchmark | Conditions | Result | Reported by | Revision | Date |
|---|---|---|---|---|---|
| TIGER-Lab/MMLU-Pro | Task mmlu_proMetric mmlu_proComparison conditions not established | 30.84 | EvalEval Reported by a third party |
Evaluated revision not stated | 2026-06-30 |
SAVRN's Notes on Mistral-7B-Instruct-v0.2
Pick this version over its predecessor for the window: 32,768 tokens of context against 8k in v0.1, with the RoPE base raised to 1e6, on a 7.2 billion parameter instruct model. At 16-bit the bfloat16 weights are 14.5 GB and the memory need is 17.4 GB. At $1.85 per hour on demand, the cheapest card we list, a single 192 GB MI300X, leaves most of its memory free for long prompts and batch. Quantized to 8-bit it needs 8.7 GB.
Apache 2.0 covers it, so commercial use, modification and redistribution are all on the table as long as the notices stay and significant changes are stated. Two things to check. Prompts must be wrapped in [INST] and [/INST] tokens, so your serving layer has to apply that template. And the one benchmark on file, 30.84 on MMLU-Pro, is third-party reported by EvalEval, not the publisher's number.
SAVRN's Notes on Qwen2.5-Coder-7B-Instruct
Code is the job here, generation, reasoning and fixing, on the 7B rung of a series spanning 0.5 to 32 billion parameters, trained on 5.5 trillion tokens over Qwen2.5. The 7.6B parameters need 18.3 GB at 16-bit, 9.1 GB at 8-bit and 4.6 GB at 4-bit, any of which fits a 192 GB MI300X at $1.85 per hour. We would never run it alone on that card; pack copies or sessions onto it.
The license is Apache 2.0, commercial use, modification and redistribution included, notices kept, changes stated. It is tuned from Qwen/Qwen2.5-Coder-7B, the base to start from with your own code. Confirm the window your stack honors: the configuration says 32,768 tokens, but arXiv:2309.00071 on YaRN context extension sits among its papers. Then Nscale on the Index, $0.01 in and $0.03 out per million tokens, the rate your card hour must beat before owning wins.
Questions
Which is larger, Mistral-7B-Instruct-v0.2 or Qwen2.5-Coder-7B-Instruct?
Qwen2.5-Coder-7B-Instruct (7.6B parameters) is larger than Mistral-7B-Instruct-v0.2 (7.2B parameters), by the parameter counts their publishers report.
Which is cheaper to run, Mistral-7B-Instruct-v0.2 or Qwen2.5-Coder-7B-Instruct?
At 4-bit, Mistral-7B-Instruct-v0.2 fits on 1x MI300X from $1.85 an hour and Qwen2.5-Coder-7B-Instruct on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Mistral-7B-Instruct-v0.2 commercially?
Yes. Mistral-7B-Instruct-v0.2 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
Can I use Qwen2.5-Coder-7B-Instruct commercially?
Yes. Qwen2.5-Coder-7B-Instruct is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.