SAVRN Model Hub · Comparisons
Qwen3-VL-2B-Instruct vs Rax-4.5
Qwen3-VL-2B-Instruct has 2.1B parameters and Rax-4.5 has 2.3B parameters; both are released under Apache License 2.0; at 16-bit, Qwen3-VL-2B-Instruct needs about 5.1 GB (1x MI300X from $1.85 an hour) and Rax-4.5 about 5.5 GB (1x MI300X from $1.85 an hour).
| Field | Qwen3-VL-2B-Instruct Qwen/Qwen3-VL-2B-Instruct | Rax-4.5 raxcore-dev/Rax-4.5 |
|---|---|---|
| Publisher | Qwen | RaxCore |
| Task | Image and text to text | Image and text to text |
| Modality | Image and text | Image and text |
| Parameters, as reported | 2.1B parameters | 2.3B parameters |
| Architecture | Qwen3VLForConditionalGeneration | Qwen3_5ForConditionalGeneration |
| Library | transformers | transformers |
| Context length | 262,144 tokens | 262,144 tokens |
| Repository size | 4.3 GB | 4.6 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) | 5.1 GB | 5.5 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 1.3 GB | 1.4 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 89644892e4d8 | 6ab58f792f61 |
| Downloads reported by the hub | 3.1M | 952.2k |
| 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.
Qwen3-VL-2B-Instruct
| Benchmark | Conditions | Result | Reported by | Revision | Date |
|---|---|---|---|---|---|
| tiiuae/PBench | Task averageMetric averageSetup Detection-only model combined with SAM2 to convert boxes to segmentation masks.Comparison conditions not established | 37 | Community Evals Reported by a third party |
Evaluated revision not stated | 2026-05-11 |
SAVRN's Notes on Rax-4.5
A context window of 262,144 tokens on 2.3 billion parameters, taking image and text in and producing text. At 16-bit the weights are 4.5 GB and the run needs 5.5 GB of memory; 8-bit brings that to 2.7 GB and 4-bit to 1.4 GB. Any of those fits on the cheapest card in our table, one MI300X with 192 GB at $1.85 an hour on demand, and the spare room matters: a filled context window adds memory on top of the weights.
RaxCore publishes it under Apache 2.0, so commercial use, modification and redistribution are allowed with the notices retained. The architecture is listed as Qwen3_5ForConditionalGeneration, yet the only relation on record points back to Rax-4.5 itself, so ask the publisher what base model and data sit underneath. No SAVRN Index host serves it by the token yet, so self-hosting is the only price we can quote. Released November 27, 2025 and last updated July 13, 2026.
Questions
Which is larger, Qwen3-VL-2B-Instruct or Rax-4.5?
Rax-4.5 (2.3B parameters) is larger than Qwen3-VL-2B-Instruct (2.1B parameters), by the parameter counts their publishers report.
Which is cheaper to run, Qwen3-VL-2B-Instruct or Rax-4.5?
At 4-bit, Qwen3-VL-2B-Instruct fits on 1x MI300X from $1.85 an hour and Rax-4.5 on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Qwen3-VL-2B-Instruct commercially?
Yes. Qwen3-VL-2B-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.
Can I use Rax-4.5 commercially?
Yes. Rax-4.5 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.