SAVRN Model Hub · Comparisons
Qwen3-Reranker-4B vs Qwen3-VL-Reranker-2B
Qwen3-Reranker-4B has 4B parameters and Qwen3-VL-Reranker-2B has 2.1B parameters; both are released under Apache License 2.0; at 16-bit, Qwen3-Reranker-4B needs about 9.7 GB (1x MI300X from $1.85 an hour) and Qwen3-VL-Reranker-2B about 5.1 GB (1x MI300X from $1.85 an hour).
| Field | Qwen3-Reranker-4B Qwen/Qwen3-Reranker-4B | Qwen3-VL-Reranker-2B Qwen/Qwen3-VL-Reranker-2B |
|---|---|---|
| Publisher | Qwen | Qwen |
| Task | Text ranking | Text ranking |
| Modality | Other | Other |
| Parameters, as reported | 4B parameters | 2.1B parameters |
| Architecture | Qwen3ForCausalLM | Qwen3VLForConditionalGeneration |
| Library | transformers | transformers |
| Context length | 40,960 tokens | 262,144 tokens |
| Repository size | 8.1 GB | 4.3 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) | 9.7 GB | 5.1 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 2.4 GB | 1.3 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 22e683669bc0 | 4bd860ac4f15 |
| Downloads reported by the hub | 2.5M | 1.2M |
| 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.
SAVRN's Notes on Qwen3-Reranker-4B
Second stage only. A reranker rescores the candidates an embedding search already pulled back, and every query-document pair is a full pass through 36 layers. That makes it compute per query rather than a memory problem: 8.0 GB of weights at 16-bit, 9.7 GB to run, 2.4 GB at 4-bit. On the cheapest setup we list, one 192 GB MI300X at $1.85 an hour, memory never binds; pairs scored per second does, so size for traffic. The 40,960-token context caps query plus document.
Nothing in Apache 2.0 stops you shipping it in a product; keep the notices, state significant changes, plus the patent grant. Check lineage and dates. It is derived from Qwen3-4B-Base, on our hub, and shares one paper with the series' embedding models, arXiv:2506.05176. Released June 3, 2025, files last updated April 16, 2026, ten months on; diff what changed before rebuilding a serving image.
SAVRN's Notes on Qwen3-VL-Reranker-2B
Put this one behind a retrieval system, not in front of a user. A reranker reorders the candidates an embedding search returns, and this 2.1 billion parameter model does that across text, images, screenshots and video, being derived from Qwen3-VL-2B-Instruct. It needs 5.1 gigabytes at 16-bit, 4.3 of them weights, 2.6 at 8-bit, 1.3 at 4-bit. On the cheapest host the Index lists, a 192 gigabyte MI300X at $1.85 an hour, it sits beside the embedding model and the generator: one card, the whole retrieval chain.
Apache 2.0 is the license, so the deployment has no user caps and no approval step; keep the license and NOTICE files and state significant changes. Access is open. Before committing, weigh the 262,144 token context length against the length of your query and document pairs, since a reranker scores every pair, and read arXiv 2601.04720, which describes it with its Qwen3-VL-Embedding companion.
Questions
Which is larger, Qwen3-Reranker-4B or Qwen3-VL-Reranker-2B?
Qwen3-Reranker-4B (4B parameters) is larger than Qwen3-VL-Reranker-2B (2.1B parameters), by the parameter counts their publishers report.
Which is cheaper to run, Qwen3-Reranker-4B or Qwen3-VL-Reranker-2B?
At 4-bit, Qwen3-Reranker-4B fits on 1x MI300X from $1.85 an hour and Qwen3-VL-Reranker-2B on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use Qwen3-Reranker-4B commercially?
Yes. Qwen3-Reranker-4B 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-VL-Reranker-2B commercially?
Yes. Qwen3-VL-Reranker-2B 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.