Up to date
Alibaba-NLP Apache 2.0 out 13 months ago
30.5B parameters, 3B active (8 of 128 experts per token), text in, text out.
Verified 2026-10-25 Rank #713 in LLMs · score 71.9 method v1 Sources 3 Downloads 30d 13.8k · #482 of 619 tracked
Picked by rule from 7 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Z-Image is a 6B parameter image generation model family featuring four variants optimized for different tasks.
Tongyi DeepResearch is the first fully open-source Web Agent available on GitHub, HuggingFace, and ModelScope to achieve performance on par with OpenAI's DeepResearch.
The 2 picks above, plus these 5. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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.
No API provider serves it, so its downloads are people running it themselves: 17 quantized builds.
Weights only, an estimate: leave room for the KV cache. Method
About 18 GB at INT4, 37 GB at FP8, 73 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2025-09-16), amber triangles are 7 dated articles (each links to it).
22 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.
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | Qwen3MoeForCausalLM |
| Model type | qwen3_moe |
| Params by dtype | BF16: 30,532,122,624 |
| Size band | L |
| MoE | yes |
| Experts | 128 |
| Experts per token | 8 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
Weights-only memory for Tongyi-DeepResearch-30B-A3B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 73 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 37 GB · Single high-end GPU (<80 GB) |
| INT4 (0.5 bytes/param) | 18 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.
quantized: 17finetune: 3adapter: 2
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| bartowski/Alibaba-NLP_Tongyi-DeepResearch-30B-A3B-GGUF | quantized | GGUF | — | — | — | bartowski | 22 | 77.5k |
| gabriellarson/Tongyi-DeepResearch-30B-A3B-GGUF | quantized | GGUF | — | — | apache-2.0 | gabriellarson | 13 | 32.8k |
| cyankiwi/Tongyi-DeepResearch-30B-A3B-AWQ-4bit | quantized | AWQ | 5.31B | M | apache-2.0 | cyankiwi | 4 | 21.9k |
| mradermacher/Tongyi-DeepResearch-30B-A3B-i1-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 6 | 16.5k |
| mradermacher/Tongyi-DeepResearch-30B-A3B-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 2 | 12k |
| DevQuasar/Alibaba-NLP.Tongyi-DeepResearch-30B-A3B-GGUF | quantized | GGUF | — | — | — | DevQuasar | 1 | 5.7k |
| noctrex/Tongyi-DeepResearch-30B-A3B-MXFP4_MOE-GGUF | quantized | GGUF | — | — | — | noctrex | 4 | 1.7k |
| m-i/Tongyi-DeepResearch-30B-A3B-mlx-fp16 | finetune | mlx | 30.5B | L | apache-2.0 | m-i | 1 | 1.3k |
| cyankiwi/Tongyi-DeepResearch-30B-A3B-AWQ-8bit | quantized | AWQ | 9.04B | L | apache-2.0 | cyankiwi | 4 | 1k |
| m-i/Tongyi-DeepResearch-30B-A3B-mlx-8Bit | quantized | mlx | 30.5B | L | apache-2.0 | m-i | 0 | 647 |
| mrtoots/Tongyi-DeepResearch-30B-A3B-mlx-8Bit | quantized | mlx | 30.5B | L | apache-2.0 | mrtoots | 0 | 642 |
| C10X/Tongyi-DeepResearch-30B-A3B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | C10X | 0 | 632 |
| SearchSwarm/SearchSwarm-30B-A3B | finetune | — | 30.5B | L | mit | SearchSwarm | 5 | 368 |
| abalogh/Tongyi-DeepResearch-30B-A3B-4bit | quantized | -4bit | 30.5B | L | apache-2.0 | abalogh | 0 | 324 |
| ramgpt/Tongyi-DeepResearch-30B-A3B-GGUF | quantized | GGUF | — | — | apache-2.0 | ramgpt | 0 | 296 |
| enacimie/Tongyi-DeepResearch-30B-A3B-Q4_K_M-GGUF | quantized | GGUF | — | — | apache-2.0 | enacimie | 0 | 252 |
| abalogh/Tongyi-DeepResearch-30B-A3B-bf16 | finetune | -bf16 | 30.5B | L | apache-2.0 | abalogh | 0 | 171 |
| ArtusDev/Alibaba-NLP_Tongyi-DeepResearch-30B-A3B-EXL3 | quantized | EXL3 | — | — | apache-2.0 | ArtusDev | 1 | 119 |
| lancew/Tongyi-DeepResearch-30B-A3B-FP8 | quantized | -FP8 | 30.5B | L | apache-2.0 | lancew | 0 | 166 |
| marcalc/Tongyi-DeepResearch-30B-A3B-Q8_0-GGUF | quantized | Q8_0 | — | — | apache-2.0 | marcalc | 0 | 162 |
| danyloooah/rgh | adapter | — | — | — | apache-2.0 | danyloooah | 1 | 0 |
| Kskip/Lumini | adapter | — | — | — | apache-2.0 | Kskip | 0 | 0 |
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.
Hub snapshot: 2026-10-25 · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-14 14:08 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.
