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 2026-10-26 07:27 UTC Rank #212 in LLMs · score 88.3 method v1 Sources 3 Downloads 30d 192.5k · #252 of 619 tracked
Picked by rule from 34 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 26 |
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 (64 quantized builds); a big base for builders (1,069 fine-tunes, adapters and merges).
Top Spaces running it: Sarath0x8f/Document-QA-bot ♥ 33 · stanley-00/slm-testing ♥ 9 · Noveumai/NovaEval ♥ 1 · Mohi117/nitish-sir-style-llm-demo ♥ 1
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 (2024-09-15), amber triangles are 3 dated articles (each links to it), grey ticks are repo commits.
1,133 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.
It supports a context length of 32,768 tokens.[1]
Generated 2026-10-07 15:04 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 |
|---|---|
| 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 | 2026-10-26 |
vLLM: supported · Qwen2ForCausalLM · 4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60SGLang: supported · Qwen2ForCausalLM · 60d941293e3af27a12c6eb64e2c3e2e1a17f100f
vLLM registry (4b1c04e76fe9ac3ca8b7e541dd423e7d5f38dc60) · SGLang registry (60d941293e3af27a12c6eb64e2c3e2e1a17f100f) · 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: 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.
finetune: 603quantized: 64merge: 17adapter: 449
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| mlx-community/Qwen2.5-3B-Instruct-4bit | finetune | -4bit | 3.09B | M | other | mlx-community | 2 | 478.6k |
| unsloth/Qwen2.5-3B-bnb-4bit | quantized | -bnb | 3.17B | M | other | unsloth | 4 | 182.1k |
| unsloth/Qwen2.5-3B-unsloth-bnb-4bit | quantized | -bnb | 3.16B | M | apache-2.0 | unsloth | 0 | 175.5k |
| unsloth/Qwen2.5-3B | finetune | — | 3.09B | M | other | unsloth | 6 | 136.8k |
| ehristoforu/coolqwen-3b-it | finetune | — | 3.09B | M | other | ehristoforu | 3 | 111.2k |
| bartowski/Qwen2.5-3B-GGUF | quantized | GGUF | — | — | — | bartowski | 6 | 60.2k |
| MaziyarPanahi/calme-3.1-instruct-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 5 | 34.2k |
| MaziyarPanahi/calme-3.1-baguette-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 1 | 33.6k |
| MaziyarPanahi/calme-3.2-instruct-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 3 | 33.4k |
| MaziyarPanahi/calme-3.1-qwenloi-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 2 | 33.3k |
| MaziyarPanahi/calme-3.2-baguette-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 1 | 32.4k |
| UBC-NLP/NileChat-3B | finetune | — | 3.40B | M | other | UBC-NLP | 24 | 29.6k |
| dphn/Dolphin3.0-Qwen2.5-3b | finetune | — | 3.09B | M | other | dphn | 36 | 18.6k |
| ThomasBaruzier/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | ThomasBaruzier | 1 | 19.3k |
| mradermacher/Qwen2.5-3B-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 17.7k |
| Mungert/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | Mungert | 5 | 15.5k |
| QuantFactory/Dolphin3.0-Qwen2.5-3b-GGUF | quantized | GGUF | — | — | other | QuantFactory | 5 | 13k |
| mradermacher/Qwen2.5-3B-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 11.6k |
| MaziyarPanahi/calme-3.3-instruct-3b | finetune | — | 3.09B | M | other | MaziyarPanahi | 5 | 10.5k |
| QuantFactory/Qwen2.5-3B-Instruct-GGUF | quantized | GGUF | — | — | other | QuantFactory | 4 | 10.1k |
| CriteriaPO/qwen2.5-3b-sft-10 | finetune | — | 3.09B | M | — | CriteriaPO | 0 | 9.1k |
| QuantFactory/raspberry-3B-GGUF | quantized | GGUF | — | — | other | QuantFactory | 4 | 7.5k |
| Alijeff1214/DeutscheLexAI_BGB | finetune | — | — | — | apache-2.0 | Alijeff1214 | 0 | 6.8k |
| OpceanAI/Yuuki-NxG | finetune | — | 3.09B | M | apache-2.0 | OpceanAI | 1 | 6.4k |
| arcee-ai/raspberry-3B | finetune | — | 3.09B | M | other | arcee-ai | 39 | 3.9k |
| mlx-community/Qwen2.5-3B-Instruct-8bit | finetune | -8bit | 3.09B | M | other | mlx-community | 0 | 5k |
| QuantFactory/Rombos-LLM-V2.5.1-Qwen-3b-GGUF | merge | GGUF | — | — | other | QuantFactory | 1 | 4.7k |
| JohnDoes4321/Qwen2.5-3B-openvino | finetune | — | — | — | other | JohnDoes4321 | 0 | 4.5k |
| mariokart59/Qwen-2.5-3B-AQLM-OA-EM-2Bit-2x8 | quantized | — | 1.01B | S | apache-2.0 | mariokart59 | 1 | 4.4k |
| yapeichang/Qwen2.5-3B-RM8B | finetune | — | — | — | apache-2.0 | yapeichang | 0 | 3.5k |
| EasierAI/Qwen-2.5-3B | quantized | GGUF | — | — | — | EasierAI | 0 | 3.3k |
| Lansechen/Qwen2.5-3B-Open-R1-GRPO-math-selected-default | finetune | — | 3.09B | M | — | Lansechen | 0 | 3.2k |
| Lansechen/Qwen2.5-3B-Open-R1-GRPO-math-selected-cosine-noRW | finetune | — | 3.09B | M | — | Lansechen | 0 | 3.2k |
| UBC-NLP/NileChat-3B-Base | finetune | — | 3.40B | M | other | UBC-NLP | 1 | 3.1k |
| mlx-community/Qwen2.5-3B-Instruct-bf16 | finetune | -bf16 | 3.09B | M | other | mlx-community | 1 | 2.9k |
| sunblaze-ucb/Qwen2.5-3B-Intuitor-MATH-1EPOCH | finetune | — | 3.40B | M | apache-2.0 | sunblaze-ucb | 0 | 2.9k |
| DevQuasar-13/Qwen.Qwen2.5-3B-GGUF | quantized | GGUF | — | — | — | DevQuasar-13 | 1 | 2.7k |
| samuelchristlie/VibeThinker-3B-gguf | quantized | gguf | — | — | mit | samuelchristlie | 0 | 2.7k |
| Volko76/Qwen2.5-Coder-3B-GGUF | quantized | GGUF | — | — | other | Volko76 | 0 | 2.6k |
| RedHatAI/Qwen2.5-3B-FP8-dynamic | quantized | -FP8 | 3.40B | M | apache-2.0 | RedHatAI | 1 | 2.5k |
| Date | Commit | SHA |
|---|---|---|
| 2024-09-20 | Upload ./tokenizer_config.json with huggingface_hub | 3aab1f1 |
| 2024-09-19 | Upload ./tokenizer_config.json with huggingface_hub | e4aa5ac |
| 2024-09-18 | update README & LICENSE | 4a34408 |
| 2024-09-18 | Update LICENSE | 31cceb5 |
| 2024-09-18 | Update README.md | 1dff604 |
| 2024-09-18 | Create LICENSE | 65fbb88 |
| 2024-09-17 | Update README.md | ddd8503 |
| 2024-09-17 | Update README.md | e866cb5 |
| 2024-09-17 | Update README.md | d9db0f6 |
| 2024-09-17 | Update README.md | 675011f |
| 2024-09-17 | Create README.md | 695e1e4 |
| 2024-09-16 | Upload tokenizer_config.json with huggingface_hub | c10dcf3 |
| 2024-09-15 | Upload folder using huggingface_hub | a895be0 |
| 2024-09-15 | initial commit | e0dd11c |
Hosted endpoints observed for Qwen2.5-3B, 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 | — | — | — | — | — | — | — | live | 2026-10-26 |
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: 2026-10-26 · HF API: 2026-10-26 07:18 UTC · Providers: 2026-10-26 07:26 UTC · Articles: 2026-10-26 07:27 UTC · Summary: 2026-10-07 15:04 UTC · Last verified 2026-10-26 07:27 UTC
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
