Up to date
Qwen Qwen License (qwen-research) out 2 years ago
3.09B parameters, text in, text out, 32K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #61 in LLMs · score 105.4 method v1 Sources 3 Downloads 30d 3.7M · #39 of 619 tracked
Changed this week: 23 new quantized builds on the Hub (329 in all).2026-10-24 · 18 new quantized builds on the Hub (324 in all).2026-10-24
Picked by rule from 31 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Qwen2.5 introduces dense decoder-only models pretrained on up to 18 trillion tokens, available in sizes from 0.5B to 72B parameters.
The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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 | Featherless | — | — | — | — | Oct 25 |
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; heavily run locally (329 quantized builds); a big base for builders (3,177 fine-tunes, adapters and merges).
Top Spaces running it: pliny-the-prompter/obliteratus ♥ 621 · cvachet/pdf-chatbot ♥ 382 · alexnasa/Wan2.2-Animate-ZEROGPU ♥ 365 · multimodalart/self-forcing ♥ 333
Weights only, an estimate: leave room for the KV cache. Method
About 1.9 GB at INT4, 3.7 GB at FP8, 7.4 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2024-09-17), amber triangles are 3 dated articles (each links to it), grey ticks are repo commits.
3,506 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
Qwen2.5-3B-Instruct is a 3.09B parameter causal language model designed for text generation.[1][2] It supports a context length of 32,768 tokens and can generate up to 8,192 tokens.[3]
Generated 2026-10-07 15:00 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | Qwen2ForCausalLM |
| Model type | qwen2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 3,085,938,688 |
| Size band | M |
| MoE | no |
| Context length | 32,768 (config.json max_position_embeddings) |
| Languages | en |
| Papers | arXiv:2407.10671 |
| License | Qwen License (qwen-research) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2ForCausalLM · 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.
Weights-only memory for Qwen2.5-3B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 7.4 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 3.7 GB · Edge (<4 GB) |
| INT4 (0.5 bytes/param) | 1.9 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Base model: Qwen/Qwen2.5-3B
quantized: 329merge: 56finetune: 1,670adapter: 1,451
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Qwen/Qwen2.5-3B-Instruct-AWQ | quantized | AWQ | 3.40B | M | other | Qwen | 26 | 3.6M |
| Qwen/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | Qwen | 222 | 3.2M |
| unsloth/Qwen2.5-3B-Instruct-unsloth-bnb-4bit | quantized | -bnb | 3.17B | M | apache-2.0 | unsloth | 6 | 2.4M |
| OpenGVLab/InternVL2_5-4B | merge | — | 3.71B | M | mit | OpenGVLab | 58 | 1.2M |
| unsloth/Qwen2.5-3B-Instruct | finetune | — | 3.09B | M | other | unsloth | 17 | 1M |
| unsloth/Qwen2.5-3B-Instruct-bnb-4bit | quantized | -bnb | 3.17B | M | apache-2.0 | unsloth | 13 | 660.1k |
| Tiiny/SmallThinker-3B-Preview | finetune | — | 3.40B | M | — | Tiiny | 415 | 617.7k |
| bartowski/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | bartowski | 36 | 544.6k |
| LiAutoAD/Ristretto-3B | finetune | — | 3.84B | M | apache-2.0 | LiAutoAD | 10 | 187.5k |
| Qwen/Qwen2.5-3B-Instruct-GPTQ-Int4 | quantized | GPTQ | 3.09B | M | other | Qwen | 7 | 145.2k |
| arcee-ai/Arcee-VyLinh | merge | — | 3.40B | M | apache-2.0 | arcee-ai | 46 | 119.8k |
| Ellbendls/Qwen-2.5-3b-Text_to_SQL | finetune | — | 3.09B | M | mit | Ellbendls | 12 | 50.9k |
| lmstudio-community/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 8 | 50.8k |
| BounharAbdelaziz/Qwen2.5-3B-GRPO-Math-GSM8K | finetune | — | 3.09B | M | — | BounharAbdelaziz | 1 | 31.7k |
| StanfordAIMI/CheXOne | finetune | — | 3.75B | M | cc-by-nc-4.0 | StanfordAIMI | 16 | 30.4k |
| Qwen/Qwen2.5-3B-Instruct-GPTQ-Int8 | quantized | GPTQ | 3.09B | M | other | Qwen | 9 | 24.5k |
| 5CD-AI/Vintern-3B-beta | finetune | — | 3.71B | M | — | 5CD-AI | 37 | 22.2k |
| mlc-ai/Qwen2.5-3B-Instruct-q4f16_1-MLC | quantized | — | — | — | — | mlc-ai | 2 | 23.1k |
| authormist/authormist-originality | finetune | — | 3.09B | M | mit | authormist | 21 | 22k |
| mradermacher/Qwen2.5-3B-Instruct-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 22.6k |
| legraphista/Qwen2.5-3B-Instruct-IMat-GGUF | quantized | GGUF | — | — | other | legraphista | 0 | 21.4k |
| yasserrmd/Coder-GRPO-3B | quantized | GGUF | 3.09B | M | apache-2.0 | yasserrmd | 7 | 18.5k |
| gaianet/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | gaianet | 1 | 16.5k |
| second-state/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | second-state | 0 | 15k |
| mradermacher/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 14.2k |
| SeeFlock/task-19-Qwen-Qwen2.5-3B-Instruct | adapter | — | — | — | — | SeeFlock | 0 | 13.9k |
| AI-Safeguard/Ivy-VL-llava | finetune | — | 3.80B | M | apache-2.0 | AI-Safeguard | 72 | 9.8k |
| nyu-visionx/Cambrian-S-3B | finetune | — | 3.49B | M | apache-2.0 | nyu-visionx | 1 | 11.9k |
| simplescaling/s1.1-3B | finetune | — | 3.40B | M | — | simplescaling | 1 | 11.9k |
| AndriLawrence/Qwen-3B-Intent-Microplan-v2 | adapter | GGUF | 3.09B | M | other | AndriLawrence | 4 | 10.8k |
| Unbabel/M-Prometheus-3B | finetune | — | 3.09B | M | other | Unbabel | 8 | 9.8k |
| lmstudio-community/Qwen2.5-3B-Instruct-MLX-4bit | quantized | MLX | 3.09B | M | other | lmstudio-community | 1 | 10.1k |
| QuantFactory/SmallThinker-3B-Preview-GGUF | quantized | GGUF | — | — | — | QuantFactory | 3 | 10k |
| MaziyarPanahi/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | — | MaziyarPanahi | 2 | 9.4k |
| Username6568/Qwen2.5-3B-Instruct-1_plus_1_equals_2_adapter | adapter | — | — | — | — | Username6568 | 0 | 9.2k |
| Username6568/Qwen2.5-3B-Instruct-1_plus_1_equals_3_adapter | adapter | — | — | — | — | Username6568 | 0 | 9.2k |
| zemelee/qwen2.5-jailbreak | finetune | — | 3.09B | M | apache-2.0 | zemelee | 45 | 6.8k |
| Username6568/Qwen2.5-3B-Instruct-1_plus_1_equals_4_adapter | adapter | — | — | — | — | Username6568 | 0 | 8.9k |
| mobilint/Qwen2.5-3B-Instruct | quantized | — | 311M | S | other | mobilint | 0 | 8.8k |
| JeffGreen311/eve-qwen2.5-3b-consciousness-soul | quantized | GGUF | 3.09B | M | other | JeffGreen311 | 1 | 8.6k |
| Date | Commit | SHA |
|---|---|---|
| 2024-09-25 | Update metadata: Add library_name to Transformers (#2) | aa8e725 |
| 2024-09-19 | Upload ./tokenizer_config.json with huggingface_hub | 82f42ba |
| 2024-09-18 | update README & LICENSE | 59188fd |
| 2024-09-18 | Update README.md | 92e8388 |
| 2024-09-18 | Create LICENSE | 45780f2 |
| 2024-09-18 | Update tokenizer_config.json | 9fbe4b5 |
| 2024-09-17 | Update README.md | bd0312c |
| 2024-09-17 | Upload folder using huggingface_hub | 3ca5f5b |
| 2024-09-17 | Create README.md | b235ea3 |
| 2024-09-17 | initial commit | 171de09 |
Hosted endpoints observed for Qwen2.5-3B-Instruct, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Featherless hf | Qwen/Qwen2.5-3B-Instruct | — | — | — | — | — | — | — | live | 2026-10-25 |
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.
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).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:33 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
