The story so far
Weights, every dated article, and repo commits on one rail.
1 dated article and 13 repo commits so far
Teal is the weights (2025-05-29), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.
Models › Qwen3-Reranker family
Qwen3-Reranker-8B
Qwen Apache 2.0 out 17 months ago
8.19B parameters, a query and passages in, relevance scores out, 40K tokens of context. 1 provider serves it from $0.050 per million input tokens.
Verified 2026-10-25 Rank #46 in Embeddings · score 89.9 method v1 Sources 3 Downloads 30d 120.8k · #297 of 619 tracked
Do people use it?120.8kdownloads, 30 days#297 of 619 we track · top 48%+5 likes, downloads −56% this week293 likesHugging Face2026-10-24
Where to run it
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | DeepInfra | $0.050 | — | 32K | — | Oct 24 |
Where it's used
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
Served by 1 API provider; 49 quantized builds for local use.
- As a service1DeepInfra
- On own hardware49quantized builds
- Built on159 fine-tunes · 6 adapters · 0 merges
- In apps3Hugging Face Spaces
Top Spaces running it: kaengreg/rusBEIR ♥ 6 · Ahmed-El-Zainy/rag_korean_manufacturing_docs ♥ 1
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 4.9 GB at INT4, 9.8 GB at FP8, 20 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
64 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
64 derivatives on the Hub, most of them quantized
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Benchmarks
6 rows
Benchmarks 6 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (6)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported FollowIR | 8.05 | Qwen3-Reranker-8B | model author | 2026-10-07 |
| author-reported MLDR · Embed text for search | 70.19 | Qwen3-Reranker-8B | model author | 2026-10-07 |
| author-reported MTEB · Embed text for search | 81.22 | Qwen3-Reranker-8B | model author | 2026-10-07 |
| author-reported MTEB Reranking · Rerank search results | 69.02 | Qwen3-Reranker-8B | model author | 2026-10-07 |
| author-reported MTEB Reranking · Rerank search results | 77.45 | Qwen3-Reranker-8B | model author | 2026-10-07 |
| author-reported MTEB Reranking · Rerank search results | 72.94 | Qwen3-Reranker-8B | model author | 2026-10-07 |
What the card says
1 cited sentence
Use-case evidence
8 use cases
What it is good at 8 use cases
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
- Rerank search resultsauthor-reportedAuthor reports CMTEB-R 77.45decision-model estimate p=1.00name: Rerank+2 more
- Embed text for searchauthor-reportedAuthor reports MLDR 70.19Author reports MTEB-Code 81.22tag: sentence-transformers
- Call tools and APIsstructuralchat template: <tool_call>
- Reason through hard problemsstructuralchat template: <think>
- Chat and follow instructionsstructuralchat template present
- Run on a laptop or phonestructuralfits: single-gpu
- Classify and make typed decisionsstructuraldecision-model estimate p=0.76
- Work across languagesstructuraldecision-model estimate p=1.00
Facts
21 fields
Listed in Embeddings
Facts
- Qwen3-Reranker-8B has 8.19B parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 40,960 tokens (40K). (config.json max_position_embeddings, observed 2026-10-24)
- License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
- Created 2025-05-29, last modified 2026-04-16. (HF API, observed 2026-10-24)
| Pipeline tag | text-ranking |
|---|---|
| Library | transformers |
| Architectures | Qwen3ForCausalLM |
| Model type | qwen3 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 8,188,548,096 |
| Size band | M |
| MoE | no |
| Context length | 40,960 (config.json max_position_embeddings) |
| Papers | arXiv:2506.05176 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: supported · Qwen3ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for Qwen3-Reranker-8B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 20 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 9.8 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.9 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
40 variants
Lineage
Base model: Qwen/Qwen3-8B-Base
quantized: 49finetune: 9adapter: 6
Top variants (40 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| boboliu/Qwen3-Reranker-8B-W4A16-G128 | quantized | -W4A16 | 8.24B | M | apache-2.0 | boboliu | 0 | 43.5k |
| mradermacher/Qwen3-Reranker-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 13 | 41.6k |
| galaxycore/Qwen3-Reranker-8B-MLX-4bit | quantized | MLX | 8.19B | M | apache-2.0 | galaxycore | 2 | 39k |
| PXY107kWqS/Qwen3-Reranker-8B-bnb-4bit | quantized | -bnb | 7.57B | M | apache-2.0 | PXY107kWqS | 0 | 22.3k |
| Voodisss/Qwen3-Reranker-8B-GGUF-llama_cpp | quantized | GGUF | — | — | apache-2.0 | Voodisss | 6 | 11.2k |
| furiosa-ai/Qwen3-Reranker-8B | finetune | — | 8.19B | M | apache-2.0 | furiosa-ai | 0 | 9.6k |
| mradermacher/Qwen3-Reranker-8B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 1 | 8k |
| DevQuasar/Qwen.Qwen3-Reranker-8B-GGUF | quantized | GGUF | — | — | — | DevQuasar | 0 | 6.5k |
| giladgd/Qwen3-Reranker-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | giladgd | 1 | 5.1k |
| xentauri/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | xentauri | 0 | 3.3k |
| drawais/Qwen3-Reranker-8B-AWQ-INT4 | quantized | AWQ | 8.24B | M | apache-2.0 | drawais | 1 | 2.3k |
| dean2155/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | dean2155 | 0 | 2.1k |
| huynguyendbs/Qwen3-Reranker-8B-4bit-MLX | finetune | -4bit | 8.19B | M | apache-2.0 | huynguyendbs | 0 | 1.5k |
| batiai/Qwen3-Reranker-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | batiai | 0 | 1.1k |
| camelliah/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | camelliah | 0 | 965 |
| zenlm/zen-reranker-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | zenlm | 0 | 885 |
| Horkios/Qwen3-Reranker-8B-mlx-4Bit | quantized | mlx | 8.19B | M | apache-2.0 | Horkios | 0 | 625 |
| ethicalabs/Qwen3-Reranker-8B-GGUF | quantized | GGUF | — | — | apache-2.0 | ethicalabs | 0 | 536 |
| camelliah/Qwen3-Reranker-8B-Q5_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | camelliah | 0 | 506 |
| sinjab/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | sinjab | 0 | 494 |
| redsquirrel/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | redsquirrel | 1 | 368 |
| ebasKing/Qwen3-Reranker-8B-GGUF-llama_cpp | quantized | GGUF | — | — | apache-2.0 | ebasKing | 0 | 396 |
| plated6913/Qwen3-Reranker-8B-mlx-4Bit | quantized | mlx | 8.19B | M | apache-2.0 | plated6913 | 0 | 371 |
| bsisduck/Qwen3-Reranker-8B-fp16-mlx | quantized | -fp16 | 8.19B | M | apache-2.0 | bsisduck | 0 | 310 |
| greenwich157/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | greenwich157 | 0 | 265 |
| redsquirrel/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | redsquirrel | 0 | 248 |
| aotsukiqx/Qwen3-Reranker-8B-Q6_K-GGUF | quantized | Q6_K | — | — | apache-2.0 | aotsukiqx | 0 | 233 |
| christian-winkler-th/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | christian-winkler-th | 0 | 231 |
| FukayaTakashi/Qwen3-Reranker-8B | quantized | GGUF | — | — | apache-2.0 | FukayaTakashi | 0 | 224 |
| yaraksan/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | yaraksan | 0 | 224 |
| sealad886/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | sealad886 | 0 | 207 |
| srpone/zoowork-shopranker-8b | adapter | — | 8.19B | M | apache-2.0 | srpone | 3 | 56 |
| hugger7326/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | hugger7326 | 0 | 181 |
| sinjab/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | sinjab | 0 | 178 |
| bb4d92fe-605b-4484-a749-fc18e6e2f169/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | bb4d92fe-605b-4484-a749-fc18e6e2f169 | 0 | 176 |
| sinjab/Qwen3-Reranker-8B-F16-GGUF | quantized | GGUF | — | — | apache-2.0 | sinjab | 0 | 174 |
| fuckoff11/Qwen3-Reranker-8B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | fuckoff11 | 0 | 165 |
| G37A/Qwen3-Reranker-8B-80k-squad | finetune | — | 8.19B | M | apache-2.0 | G37A | 1 | 108 |
| ExpedientFalcon/Qwen3-Reranker-8B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | ExpedientFalcon | 0 | 156 |
| hoailebads/Qwen3-Reranker-8B-VLSP-Legal-LoRA | adapter | — | — | — | apache-2.0 | hoailebads | 1 | 100 |
Repository changelog
13 commits
Repository changelog 13 commits
| Date | Commit | SHA |
|---|---|---|
| 2026-04-16 | Integrate with Sentence Transformers v5.4 (#11) | 77d193c |
| 2025-06-09 | update model type and add paper link (#4) | 5fa9408 |
| 2025-06-06 | Update README.md | d678ef8 |
| 2025-06-05 | Update README.md | 6377d20 |
| 2025-06-05 | Update README.md | 7bfde04 |
| 2025-06-05 | Update README.md | 8cd7869 |
| 2025-06-05 | Delete special_tokens_map.json | da8539f |
| 2025-06-05 | Delete added_tokens.json | 1195d2a |
| 2025-06-05 | Update README.md | d6813ab |
| 2025-06-05 | Update README.md | 5038fce |
| 2025-06-05 | Update README.md | c5f3fad |
| 2025-05-29 | Upload folder using huggingface_hub | d232ad0 |
| 2025-05-29 | initial commit | 9146df8 |
All providers
1 provider · 1 endpoint
Where to run 1 providers · 1 endpoints
Hosted endpoints observed for Qwen3-Reranker-8B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting. Cheapest observed input price: $0.05/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| DeepInfra deepinfra | Qwen/Qwen3-Reranker-8B | 32K | — | $0.05 | — | — | — | — | live | 2026-10-24 |
Articles and coverage
4 pieces
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
Analysis4
reranker.ukabout the Qwen3 Reranker family762 words
Qwen3-Reranker offers Apache 2.0 rerankers in 0.6B, 4B, and 8B sizes with a 32K context and 100+ languages. The 4B and 8B variants lead on MTEB-R and FollowIR suites across English, Chinese, multilingual, long-document, and code domains. Engineers should validate on their own labels before replacing bge or hosted APIs, noting that GPU resources are required for comfortable operation.written, reviewed, cited
- The 8B variant leads on Chinese, multilingual, long-document, and code tasks by at most 1.5 points.
- The raw transformers path requires Qwen's exact prompt template, otherwise quality drops without an error.
- The 4B model achieves the best FollowIR score in Qwen's table among the available sizes.
agentset.ai459 words
Qwen3 Reranker 8B is an open-source cross-encoder reranker built on the Qwen foundation model released on 2025-06-06. It achieves 69.02 on MTEB-R and 81.22 on MTEB-Code benchmarks with 3300 total evaluations. The model costs $0.050 per 1M tokens and supports 32K context across 100+ languages.written, reviewed, cited
aichina.news1,143 words
Qwen3-Reranker-8B is a text reranking model designed to re-order search results for maximum accuracy in RAG workflows. It supports over 100 languages and handles a context length of 32k tokens. The Qwen3-Embedding-8B model achieves a score of 70.58 on the MTEB multilingual leaderboard as of 5 June 2025. Using specific instructions typically yields a 1% to 5% improvement for most downstream tasks.written, reviewed, cited
- In multilingual contexts, instructions should be written in English because most training instructions were English-based.
- Transformers versions earlier than 4.45.0 may encounter a KeyError when loading the model.
blogs.novita.ai2025-06-131,233 words
Novita AI releases the Qwen 3 Reranker 8B model, which uses a cross-encoder setup to reorder documents with a 32K sequence length. The model achieves MTEB-R scores of 69.02, CMTEB-R scores of 77.45, and MTEB-Code scores of 81.22. Developers access the model via the Novita AI API platform.written, reviewed, cited
- Novita AI provides the free bge-m3 model alongside its paid Qwen 3 Reranker 8B offering.
- The Qwen3-Reranker-8B model achieves an MTEB-Code score of 81.22, outperforming other listed rerankers.
- Developers can refer friends to receive up to $500 in total LLM API credits as a reward.
Papers
1 paper
Papers 1
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Similar models
3 models
Similar models
- Qwen3-VL-Embedding-8B · 8.14B same size band (M), permissive license
- Qwen3-Embedding-8B · 7.57B same size band (M), more providers, permissive license
- bge-multilingual-gemma2 · 9.24B closest by size and task
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:35 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:01 UTC · Last verified 2026-10-25
Follow Qwen3-Reranker-8B
Watch this model
- /feeds/models/Qwen--Qwen3-Reranker-8B.xml(RSS 2.0) · every change to Qwen3-Reranker-8B
- /feeds/families/Qwen--Qwen3-Reranker.xml(RSS 2.0) · every change across the Qwen3-Reranker family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.