SAVRN Model Hub · Comparisons
Qwen2.5-1.5B vs Qwen3-1.7B
Qwen2.5-1.5B has 1.5B parameters and Qwen3-1.7B has 2B parameters; both are released under Apache License 2.0; at 16-bit, Qwen2.5-1.5B needs about 3.7 GB (1x MI300X from $1.85 an hour) and Qwen3-1.7B about 4.9 GB (1x MI300X from $1.85 an hour).
| Field | Qwen2.5-1.5B Qwen/Qwen2.5-1.5B | Qwen3-1.7B Qwen/Qwen3-1.7B |
|---|---|---|
| Publisher | Qwen | Qwen |
| Task | Text generation | Text generation |
| Modality | Text | Text |
| Parameters, as reported | 1.5B parameters | 2B parameters |
| Architecture | Qwen2ForCausalLM | Qwen3ForCausalLM |
| Library | transformers | transformers |
| Context length | 131,072 tokens | 40,960 tokens |
| Repository size | 3.1 GB | 4.1 GB |
| Artifact formats | safetensors | safetensors |
| License | apache-2.0 | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 3.7 GB | 4.9 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0.9 GB | 1.2 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 8faed761d45a | 70d244cc86cc |
| Downloads reported by the hub | 880.3k | 3.7M |
| Last observed | 2026-09-18 | 2026-10-13 |
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.
Qwen2.5-1.5B
| Benchmark | Conditions | Result | Reported by | Revision | Date |
|---|---|---|---|---|---|
| TIGER-Lab/MMLU-Pro | Task mmlu_proMetric mmlu_proComparison conditions not established | 32.1 | EvalEval Reported by a third party |
Evaluated revision not stated | 2026-06-30 |
SAVRN's Notes on Qwen2.5-1.5B
3.7 GB at 16-bit, 0.9 GB at 4-bit, 1.5B parameters, a 131,072-token context, 880.3k downloads a month: this is the smallest dense entry in the Qwen2.5 family and it is a base model, which the download count tends to hide. The cheapest Index rental is one MI300X at $1.85 an hour, but at this size the sensible hosts are a laptop GPU or a sidecar slot on a server you already run.
Apache 2.0 allows commercial fine-tuning and redistribution with notices kept. The one number on file is 32.1 on MMLU-Pro, third-party reported — respectable for the size but a base-model figure, not a chat figure. If you want a ready assistant, the Instruct variant one slot over in similar models is the correct pick; if you are doing continued pretraining or a domain-adaptation fine-tune, start here and save the alignment tax for later.
SAVRN's Notes on Qwen3-1.7B
Two billion parameters, 4.9 GB at 16-bit, 1.2 GB at 4-bit: this is a tenant, not a workload. The cheapest listed setup — one MI300X at $1.85 an hour — is never sized for it; it rides in the spare memory of a card already earning its keep, and 3.2M downloads a month say most operators treat it that way. The 40,960-token context is the figure to design against: the page's 20% margin covers a short context, and a full window needs more than the 1.2 GB headline.
Apache 2.0 permits commercial use, modification and redistribution, with notices kept and changes stated. The listing derives this checkpoint from Qwen/Qwen3-1.7B-Base, so teams doing their own instruction tuning start from the base, not here. The weights are 4.1 GB of safetensors across 12 files, last updated July 2025. No Index host sells it by the token; at this footprint, none should — the card hour is the whole price.
Questions
Which is larger, Qwen2.5-1.5B or Qwen3-1.7B?
Qwen3-1.7B (2B parameters) is larger than Qwen2.5-1.5B (1.5B parameters), by the parameter counts their publishers report.
Which is cheaper to run, Qwen2.5-1.5B or Qwen3-1.7B?
At 4-bit, Qwen2.5-1.5B fits on 1x MI300X from $1.85 an hour and Qwen3-1.7B on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Qwen2.5-1.5B commercially?
Yes. Qwen2.5-1.5B 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 Qwen3-1.7B commercially?
Yes. Qwen3-1.7B 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.