Up to date
Qwen Qwen License (qwen) out 21 months ago
73.4B parameters, images and text in, text out, 125K tokens of context. 2 providers serve it from $0.80 per million input tokens.
Verified 2026-10-25 Rank #94 in LLMs · score 99.6 method v1 Sources 5 Downloads 30d 96.1k · #319 of 619 tracked
Changed this week: 1 provider no longer listed: Featherless.2026-10-23
Picked by rule from 32 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 1 pick above, plus these 24. 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.
2 providers with a live price, from $0.80 to $1.01 per million input tokens; the median is $0.91. One dot per provider, cheapest in teal.
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 2 API providers; 33 quantized builds for local use.
Top Spaces running it: smolagents/computer-agent ♥ 983 · linoyts/Qwen-Image-Edit-2511-AnyPose ♥ 415 · linoyts/Qwen-Image-Edit-2511-Fast ♥ 272 · linoyts/Qwen-Image-Edit-next-scene ♥ 164
Weights only, an estimate: leave room for the KV cache. Method
About 44 GB at INT4, 88 GB at FP8, 176 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Claim-matched tests, graded by code. Receipts, not a score.
{
"providers": [
{
"provider": "parasail",
"endpoint_model": "parasail-qwen25-vl-72b-instruct",
"source_url": "https://api.parasail.io/v1/chat/completions",
"ttft_ms_median": 489,
"ttft_any_ms_median": 489,
"tokens_per_s_median": 49.3,
"runs": [
{
"ttft_s": 0.827,
"ttft_any_s": 0.827,
"tokens_per_s": 40.1,
"latency_s": 3.86,
"tokens_out": 155,
"reasoning_chars": 0,
"finish": "stop"
},
{
"ttft_s": 0.454,
"ttft_any_s": 0.454,
"tokens_per_s": 49.3,
"latency_s": 3.27,
"tokens_out": 161,
"reasoning_chars": 0,
"finish": "stop"
},
{
"ttft_s": 0.489,
"ttft_any_s": 0.489,
"tokens_per_s": 52.6,
"latency_s": 2.89,
"tokens_out": 152,
"reasoning_chars": 0,
"finish": "stop"
}
]
}
],
"note": "Single client, wall clock; tok/s counts every output token including hidden reasoning."
} {
"fields_matched": 79,
"fields_total": 80,
"cost_per_1000_usd": 0.0576,
"parsed": true,
"per_item": [
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": false,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
},
{
"street": true,
"city": true,
"postcode": true,
"country": true
}
],
"sample": [
{
"street": "1600 Amphitheatre Parkway",
"unit": null,
"city": "Mountain View",
"region": "CA",
"postcode": "94043",
"country": "US"
},
{
"street": "10 Downing Street",
"unit": null,
"city": "London",
"region": null,
"postcode": "SW1A 2AA",
"country": "GB"
},
{
"street": "Rua Augusta 274",
"unit": null,
"city": "Lisbon",
"region": null,
"postcode": "1100-053",
"country": "PT"
}
],
"fixture": "receipts.addresses.v1"
} Weights, every dated article, and repo commits on one rail.
Teal is the weights (2025-01-27), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
78 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.
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 | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | Qwen2_5_VLForConditionalGeneration |
| Model type | qwen2_5_vl |
| Dtype | bfloat16 |
| Params by dtype | BF16: 73,410,777,344 |
| Size band | XL |
| MoE | no |
| Context length | 128,000 (config.json max_position_embeddings) |
| Papers | arXiv:2309.00071, arXiv:2409.12191, arXiv:2308.12966 |
| License | Qwen License (qwen) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2_5_VLForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2_5_VLForConditionalGeneration · 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-VL-72B-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 176 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 88 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 44 GB · Single high-end GPU (<80 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 33finetune: 30adapter: 15
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Qwen/Qwen2.5-VL-72B-Instruct-AWQ | quantized | AWQ | 73.7B | XL | other | Qwen | 79 | 2.2M |
| RedHatAI/Qwen2.5-VL-72B-Instruct-FP8-dynamic | quantized | -FP8 | 73.4B | XL | apache-2.0 | RedHatAI | 15 | 371.1k |
| unsloth/Qwen2.5-VL-72B-Instruct-bnb-4bit | quantized | -bnb | 73.4B | XL | other | unsloth | 15 | 85.7k |
| unsloth/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | other | unsloth | 11 | 81.5k |
| PointerHQ/Qwen2.5-VL-72B-Instruct-Pointer-AWQ | quantized | AWQ | 73.7B | XL | other | PointerHQ | 8 | 74.6k |
| lmstudio-community/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | other | lmstudio-community | 2 | 57.3k |
| Benasd/Qwen2.5-VL-72B-Instruct-AWQ | quantized | AWQ | 73.4B | XL | other | Benasd | 6 | 55.5k |
| Mungert/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | other | Mungert | 12 | 51.2k |
| parasail-ai/Qwen2.5-VL-72B-Instruct-FP8-Dynamic | quantized | -FP8 | 73.4B | XL | apache-2.0 | parasail-ai | 2 | 44.3k |
| ggml-org/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | apache-2.0 | ggml-org | 1 | 36.7k |
| bartowski/Qwen_Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | other | bartowski | 5 | 31.6k |
| unsloth/Qwen2.5-VL-72B-Instruct-unsloth-bnb-4bit | quantized | -bnb | 73.4B | XL | other | unsloth | 8 | 28.1k |
| mlx-community/Qwen2.5-VL-72B-Instruct-4bit | finetune | -4bit | 73.4B | XL | other | mlx-community | 10 | 21.3k |
| samgreen/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | other | samgreen | 1 | 15.7k |
| RedHatAI/Qwen2.5-VL-72B-Instruct-quantized.w4a16 | quantized | -quant | 73.4B | XL | apache-2.0 | RedHatAI | 8 | 14.3k |
| RedHatAI/Qwen2.5-VL-72B-Instruct-quantized.w8a8 | quantized | -quant | 73.4B | XL | apache-2.0 | RedHatAI | 1 | 13.5k |
| mradermacher/Qwen2.5-VL-72B-Instruct-i1-GGUF | quantized | GGUF | — | — | — | mradermacher | 0 | 11.9k |
| xlangai/OpenCUA-72B | finetune | — | 73.4B | XL | mit | xlangai | 8 | 5.6k |
| DevQuasar/Qwen.Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | — | DevQuasar | 0 | 4.8k |
| mlx-community/Qwen2.5-VL-72B-Instruct-3bit | finetune | — | 73.4B | XL | other | mlx-community | 5 | 3.4k |
| GeorgyGUF/INFRL-Qwen2.5-VL-72B-Preview-ggufs-fully-quantized | quantized | gguf | — | — | apache-2.0 | GeorgyGUF | 0 | 3.7k |
| mlx-community/Qwen2.5-VL-72B-Instruct-8bit | finetune | -8bit | 73.4B | XL | other | mlx-community | 2 | 3.3k |
| Hcompany/Holo1.5-72B | finetune | — | 73.4B | XL | cc-by-nc-4.0 | Hcompany | 27 | 1.1k |
| mradermacher/Qwen2.5-VL-72B-Instruct-GGUF | quantized | GGUF | — | — | — | mradermacher | 0 | 2.4k |
| Benasd/Qwen2.5-VL-72B-Instruct-AWQ-fix | quantized | AWQ | 73.7B | XL | other | Benasd | 1 | 2.1k |
| AbteeXAILab/lumynax-multimodal-qwen25-vl-72b-instruct-gguf | quantized | gguf | — | — | other | AbteeXAILab | 0 | 2k |
| ig1/Qwen2.5-VL-72B-Instruct-FP8-Dynamic | quantized | -FP8 | 73.4B | XL | other | ig1 | 0 | 1.7k |
| unsloth/Qwen2.5-VL-72B-Instruct | finetune | — | 73.4B | XL | other | unsloth | 1 | 1.7k |
| chancharikm/qwen2.5-vl-72b-cam-motion | finetune | — | 73.4B | XL | other | chancharikm | 1 | 1.5k |
| mlx-community/Qwen2.5-VL-72B-Instruct-6bit | finetune | — | 73.4B | XL | other | mlx-community | 1 | 1.4k |
| GeorgyGUF/Qwen2.5-VL-72B-Instruct-q8_0-with-bf16-output-and-bf16-embedding.gguf | quantized | q8_0 | — | — | other | GeorgyGUF | 0 | 1.5k |
| baichuan-inc/BaichuanMed-OCR-72B | finetune | — | 73.4B | XL | other | baichuan-inc | 8 | 949 |
| anongecko/Qwen2.5-VL-72B-Instruct-heretic-FP8 | quantized | -FP8 | 73.4B | XL | other | anongecko | 1 | 1.2k |
| mlx-community/Qwen2.5-VL-72B-Instruct-bf16 | finetune | -bf16 | 73.4B | XL | other | mlx-community | 0 | 1.2k |
| GeorgyGUF/Qwen2.5-VL-72B-Instruct-bf16.gguf | quantized | -bf16 | — | — | other | GeorgyGUF | 1 | 826 |
| coder3101/Qwen2.5-VL-72B-Instruct-heretic | finetune | — | 73.4B | XL | other | coder3101 | 2 | 681 |
| mukulp/Qwen2.5-VL-72B-Instruct-bf16 | finetune | -bf16 | 73.4B | XL | other | mukulp | 0 | 681 |
| TIGER-Lab/VL-Reasoner-72B | finetune | — | 73.4B | XL | apache-2.0 | TIGER-Lab | 3 | 485 |
| Geraldxm/Qwen2.5-VL-72B-Instruct-eagle3-sgl | finetune | — | 1.43B | S | mit | Geraldxm | 1 | 505 |
| AI45Research/SafeWork-R1 | finetune | — | 73.4B | XL | apache-2.0 | AI45Research | 6 | 241 |
| Date | Commit | SHA |
|---|---|---|
| 2025-06-06 | remove self referencing base model (#32) | 89c8620 |
| 2025-03-23 | Update generation_config.json | cd3b627 |
| 2025-03-21 | Update generation_config.json | 71dfcaa |
| 2025-03-07 | Update tokenizer_config.json | d91279c |
| 2025-02-15 | Update preprocessor_config.json | 5d8e171 |
| 2025-01-27 | Update README.md | 699299c |
| 2025-01-27 | Update README.md | b937ff8 |
| 2025-01-27 | Update README.md | cd4c8c1 |
| 2025-01-27 | Create LICENSE | 95a566e |
| 2025-01-27 | Create README.md | 98b8649 |
| 2025-01-27 | Upload folder using huggingface_hub | a1a5805 |
| 2025-01-27 | initial commit | 73979ce |
Hosted endpoints observed for Qwen2.5-VL-72B-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. Cheapest observed input price: $0.80/M.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Parasail parasail, openrouter | parasail-qwen25-vl-72b-instruct | 125K | fp8 | $0.80 | $1.00 | — | — | — | serverless | 2026-10-24 |
| OVHcloud hf | Qwen2.5-VL-72B-Instruct | 32K | — | $1.01 | $1.01 | 29.5 | 1170 ms | no | 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.