Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › Qwen3.8 family

Qwen3.8-2.4T-A95B

Hugging Face ↗ Follow changes

Qwen Qwen License (qwen3.8-max) out 3 months ago

2.45T parameters, 95B active (10 of 512 experts per token), text in, text out, 256K tokens of context. 7 providers serve it from $2.00 per million input tokens.

Verified 2026-10-25 Rank #118 in LLMs · score 96.6 method v1 Sources 4 Downloads 30d 25.4k · #439 of 619 tracked

Can I run it?1.47TB at INT4Estimateneeds a multi-GPU nodemethodCommunity: Qwen3.8 2.4T A95B in-depth review: 2.4 trillion parameter open source weight, a milestone for Alibaba Tongyi Qianwen – AI Dash →
What does it cost?$2.00/1M incheapest of 6Alibaba2026-10-24
Can I use it?Qwen License (qwen3.8-max)Free below a usage thresholdlicence file2026-10-24
Do people use it?25.4kdownloads, 30 days#439 of 619 we track · below the median+12 likes, downloads −3% this week1,327 likesHugging Face2026-10-24

Worth reading

Picked by rule from 31 tracked pieces: the lab's own post, the most shared, a field report, the latest.

25 best of 31 tracked

The 3 picks above, plus these 22. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. ArticleQwen3.8-Flash-Nextqwen.ai · 2026-08-26 · official
  2. ArticleQwen3.8-Max: A New Bar for Coding and Coworkqwen.ai · 2026-08-03 · official
  3. ArticleQwen3.8 27B scores 52 on Artificial Analysisartificialanalysis.ai · 2026-08-17
  4. ArticleQwen3.8 Max now ranked as the best overall model by agentic indexartificialanalysis.ai · 2026-08-06
  5. CodeShow HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/sgithub.com · 2026-09-01
  6. ArticleBenchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapsesquesma.com · 2026-08-26
  7. ArticleRun Qwen3.8 27B locally: real numbers from my Mac Studioterminalbytes.com · 2026-08-28
  8. CodeRun QWEN3.8 27B on 16gb Nvidia GPUsgithub.com · 2026-09-17
  9. ArticleQwen3.8-Flash-Next Intelligence, Performance and Price Analysisartificialanalysis.ai · 2026-08-27
  10. DiscussionQwen3.8-27Btwitter.com · 2026-08-14
  11. ArticleQwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiencyqwen.ai · 2026-08-26 · official
  12. ArticleQwen3.8-Maxqwencloud.com · 2026-09-02
  13. ArticleQwen3.8 27B at 256K: 50 TPS on a 24 GB GPUpiszczek.pl · 2026-08-17
  14. CodeQwen3.8-Flash-Next Technical Report [pdf]github.com · 2026-08-26
  15. CodeQwen3.8-27B make medium the default effort level instead of xhighgithub.com · 2026-08-19
  16. CodeI found why Qwen3.8-27B was 14× slower on my 4070 Supergithub.com · 2026-08-23
  17. ArticleServe Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72 | NVIDIA Technical Blogdeveloper.nvidia.com · 2026-08-12
  18. DiscussionQwen3.8 Fetches Weird URLsnews.ycombinator.com · 2026-08-20 · 8 HN pts
  19. DiscussionQwen3.8-Maxtwitter.com · 2026-08-03
  20. ArticleQwen3.8 2.4T A95B vs gpt-oss-120b (High)artificialanalysis.ai · undated
  21. ArticleWhat Is Qwen3.8-2.4T-A95B? Alibaba's 2.4T MoEkie.ai · 2026-08-17
  22. ArticleQwen3.8-2.4T-A95B · LLM Model Review | CrucibleMarkcruciblemark.com · 2026-08-15

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

6 providers from $2.00

6 providers with a live price, from $2.00 to $2.00 per million input tokens; the median is $2.00. One dot per provider, cheapest in teal.

$0.010$0.10$1.00$10.00Novita: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingTogether AI: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingDeepInfra: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingAlibaba: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bSiliconFlow: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bVenice: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bcheapest $2.00median $2.00
$0.010$0.10$1.00$10.00Novita: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingTogether AI: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingDeepInfra: $2.00 in / $6.00 out per 1M · observed 2026-10-25 · https://huggingface.co/api/models/Qwen/Qwen3.8-2.4T-A95B?expand[]=inferenceProviderMappingAlibaba: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bSiliconFlow: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bVenice: $2.00 in / $6.00 out per 1M · observed 2026-10-24 · https://openrouter.ai/qwen/qwen3.8-2.4t-a95bcheapest $2.00 · median $2.00 (white tick)
#Cheapest 5In /1MOut /1MContextQuantSeen
1 Alibaba $2.00 $6.00 977K unknown Oct 24
2 DeepInfra $2.00 $6.00 256K — Oct 25
3 Novita $2.00 $6.00 977K unknown Oct 25
4 SiliconFlow $2.00 $6.00 1M fp8 Oct 24
5 Together AI $2.00 $6.00 986K unknown Oct 25

All 7 providers →

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.

Widely served (7 API providers); 33 quantized builds for local use.

  • As a service7Alibaba, DeepInfra, Fireworks AI, Novita +3
  • On own hardware33quantized builds
  • Built on77 fine-tunes · 0 adapters · 0 merges
  • In apps23Hugging Face Spaces

Top Spaces running it: FineEnvs/simulation-rl-environments ♥ 12 · Svngoku/jupyter2agent ♥ 8 · pollen-robotics/reachy-mini-text-to-motion ♥ 6 · llamameta/glm5.3-xliberated-free ♥ 4

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Too big for one 8×80 GB node at any precision

About 1.47 TB at INT4, 2.94 TB at FP8, 5.87 TB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 1.47 TB of weights (estimate; KV cache and activations extra)INT41.47 TBFP8: about 2.94 TB of weights (estimate; KV cache and activations extra)FP82.94 TBBF16: about 5.87 TB of weights (estimate; KV cache and activations extra)BF165.87 TB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 1.47 TB of weights (estimate; KV cache and activations extra)INT41.47 TBFP8: about 2.94 TB of weights (estimate; KV cache and activations extra)FP82.94 TBBF16: about 5.87 TB of weights (estimate; KV cache and activations extra)BF165.87 TB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

Most coverage came in the first two weeks

Teal is the weights (2026-08-08), amber triangles are 9 dated articles (each links to it), grey ticks are repo commits.

SepOct2026-08-12: 3 repo commits2026-08-11: 2 repo commits2026-08-10: 1 repo commit2026-08-09: 1 repo commitQwen/Qwen3.8-2.4T-A95B · 2026-08-08Qwen3.8 2.4T A95B: Intelligence, Performance and Price Analysis · 2026-08-14I Rented a 96 GB GPU and Took Uncensored Qwen3.8 From 44 to 125 tok/s · 2026-08-24Serve Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72 | NVIDIA Technical Blog · 2026-08-12Qwen3.8 Fetches Weird URLs · 2026-08-20What Is Qwen3.8-2.4T-A95B? Alibaba's 2.4T MoE · 2026-08-17Qwen3.8-2.4T-A95B · LLM Model Review | CrucibleMark · 2026-08-15Qwen3.8 2.4T A95B in-depth review: 2.4 trillion parameter open source weight, a milestone for Alibaba Tongyi Qianwen – AI Dash · 2026-08-15Qwen3.8-2.4T-A95B Benchmarks vs Opus 4.8 and GPT-5.6 Sol | MindStudio · 2026-08-13Weights 2026-08-08 · 2026-08-08Weights 2026-08-08now · 7 providers
SepOct2026-08-12: 3 repo commits2026-08-11: 2 repo commits2026-08-10: 1 repo commit2026-08-09: 1 repo commitQwen/Qwen3.8-2.4T-A95B · 2026-08-08Qwen3.8 2.4T A95B: Intelligence, Performance and Price Analysis · 2026-08-14I Rented a 96 GB GPU and Took Uncensored Qwen3.8 From 44 to 125 tok/s · 2026-08-24Serve Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72 | NVIDIA Technical Blog · 2026-08-12Qwen3.8 Fetches Weird URLs · 2026-08-20What Is Qwen3.8-2.4T-A95B? Alibaba's 2.4T MoE · 2026-08-17Qwen3.8-2.4T-A95B · LLM Model Review | CrucibleMark · 2026-08-15Qwen3.8 2.4T A95B in-depth review: 2.4 trillion parameter open source weight, a milestone for Alibaba Tongyi Qianwen – AI Dash · 2026-08-15Qwen3.8-2.4T-A95B Benchmarks vs Opus 4.8 and GPT-5.6 Sol | MindStudio · 2026-08-13Weights 2026-08-08 · 2026-08-08Weights 2026-08-08now · 7 providers

What people built on it

40 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

40 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.

33 quantized (Hugging Face model tree)7 fine-tunes (Hugging Face model tree)33 quantized7 fine-tunes
33 quantized (Hugging Face model tree)7 fine-tunes (Hugging Face model tree)33 quantized7 fine-tunes
GGUF unsloth/Qwen3.8-2.4T-A95B-GGUF for llama.cpp, Ollama, LM Studio 38.8k downloads MLX pipenetwork/Qwen3.8-2.4T-A95B-MLX-reap75-4bit for Apple silicon 7.2k downloads FP8 Qwen/Qwen3.8-2.4T-A95B-FP8 for vLLM, SGLang 36.1k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

10 rows

Benchmarks 10 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (10)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.
BenchmarkValueSetting / notesReported byDate
author-reported AndroidBench75.1Qwen3.8-Maxmodel author2026-10-07
author-reported DeepSWE · Write code with agents56.6Qwen3.8-Maxmodel author2026-10-07
author-reported FrontierSWE · Write code with agents73.5Qwen3.8-Maxmodel author2026-10-07
author-reported MLS-Bench-Lite41Qwen3.8-Maxmodel author2026-10-07
author-reported NL2Repo · Write code with agents55.9Qwen3.8-Maxmodel author2026-10-07
author-reported PaperBench93Qwen3.8-Maxmodel author2026-10-07
author-reported QwenQoderBench58.4Qwen3.8-Maxmodel author2026-10-07
author-reported SWE-bench · Write code with agents80.7Qwen3.8-Maxmodel author2026-10-07
author-reported SWE-bench Pro · Write code with agents67.7Qwen3.8-Maxmodel author2026-10-07
author-reported Terminal-Bench · Write code with agents86.6Qwen3.8-Maxmodel author2026-10-07

What the card says

1 cited sentence

Summary generated, cited

The architecture utilizes 512 experts and supports a native context length of 262,144 tokens.[1][2]

  1. num experts
  2. “Context Length: 262,144 natively”

Generated 2026-10-07 15:01 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

5 use cases

What it is good at 5 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.

Facts

23 fields

Listed in LLMs · MoE

Facts

Pipeline tagtext-generation
Librarytransformers
ArchitecturesQwen3_5MoeForCausalLM
Model typeqwen3_5_moe_text
Dtypebfloat16
Params by dtypeBF16: 2,446,182,725,504
Size bandXXL
MoEyes
Experts512
Experts per token10
Context length262,144 (config.json max_position_embeddings)
LicenseQwen License (qwen3.8-max) (other)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3_5MoeForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3_5MoeForCausalLM · 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 Qwen3.8-2.4T-A95B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)5.87 TB · Cluster (≥640 GB)
FP8 (1 byte/param)2.94 TB · Cluster (≥640 GB)
INT4 (0.5 bytes/param)1.47 TB · Cluster (≥640 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.

All variants

40 variants

Lineage

Original model Family: Qwen/Qwen3.8 40 derivatives on the Hub

quantized: 33finetune: 7

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
Qwen/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other Qwen 255 36.1k
unsloth/Qwen3.8-2.4T-A95B-GGUF quantized GGUF — — other unsloth 122 38.8k
nvidia/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 1.26T XXL other nvidia 24 13.9k
RadixArk/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 1.26T XXL other RadixArk 6 14.2k
Inferact/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 1.25T XXL other Inferact 2 14.2k
inference-optimization/Qwen3.8-1.0B-A0.6B finetune — 966M S mit inference-optimization 7 11.6k
pipenetwork/Qwen3.8-2.4T-A95B-MLX-reap75-4bit quantized MLX 641B XXL apache-2.0 pipenetwork 2 7.2k
pipenetwork/Qwen3.8-2.4T-A95B-MLX-reap60-3bit quantized MLX 998B XXL apache-2.0 pipenetwork 1 5.6k
a4lg/Qwen3.8-2.4T-A95B-MTP-ONLY-GGUF quantized GGUF — — other a4lg 2 5k
pipenetwork/Qwen3.8-2.4T-A95B-MLX-reap50-3bit quantized MLX 1.23T XXL apache-2.0 pipenetwork 0 3.2k
RedHatAI/Qwen3.8-2.4T-A95B-NVFP4-REAP-25 quantized NVFP4 1.85T XXL — RedHatAI 10 2.5k
anemll/Qwen3.8-2.4T-A95B-FlashMoE-UD-Q1_0 quantized GGUF — — other anemll 5 2.3k
RedHatAI/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 1.38T XXL — RedHatAI 1 2.2k
nota-ai/Qwen3.8-2.4T-A95B-Nota-NVFP4-Global-Pruned-40 quantized NVFP4 1.48T XXL other nota-ai 23 1.1k
kernelpool/Qwen3.8-2.4T-A95B-3bit-UVMAX quantized — 2.42T XXL other kernelpool 2 2k
RedHatAI/Qwen3.8-2.4T-A95B-NVFP4-REAP-50 quantized NVFP4 1.26T XXL — RedHatAI 4 1.8k
Inferact/Qwen3.8-2.4T-A95B-MXFP4 quantized MXFP4 1.25T XXL other Inferact 0 1.6k
RedHatAI/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other RedHatAI 0 1.5k
RedHatAI/Qwen3.8-2.4T-A95B-NVFP4-FP8 quantized NVFP4 1.38T XXL — RedHatAI 7 1.1k
dudeman2512/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 2.45T XXL apache-2.0 dudeman2512 0 1.3k
modal-labs/Qwen3.8-2.4T-A95B-NVFP4 quantized NVFP4 1.26T XXL — modal-labs 1 1k
Blackfrost-AI/Qwen-3.8-2.4T-A95B-Derisked-IQ1_0-GGUF quantized GGUF — — other Blackfrost-AI 4 579
dudeman2512/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL apache-2.0 dudeman2512 0 564
m8than/Qwen38-2.4T-GG quantized — 349B XXL other m8than 0 563
dudeman2512/Qwen3.8-2.4T-A95B-int4 quantized int4 2.45T XXL apache-2.0 dudeman2512 0 500
vs4vijay/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other vs4vijay 0 469
DevQuasar/Qwen.Qwen3.8-2.4T-A95B-GGUF quantized GGUF — — — DevQuasar 0 440
LilacGhost10/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other LilacGhost10 0 437
iceDonkey/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other iceDonkey 0 425
mmnga-o/Qwen3.8-REAP30-Width50-UD-gguf quantized gguf — — other mmnga-o 4 219
neopolita/Qwen3.8-84B-A53B-Niwaki-v2-experimental-mlx quantized mlx 84.0B XL other neopolita 0 386
PYTHAI/Qwen3.8-2.4T-A95B-fork finetune — — — other PYTHAI 0 254
ckoh04/Qwen3.8-2.4T-A95B-NVFP4-active-slice quantized NVFP4 72.9B XL other ckoh04 0 163
Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-UD-IQ3_XXS quantized GGUF — — other Blackfrost-Research 2 34
tiny-random/qwen3.8-moe finetune — 12.9M S — tiny-random 0 59
yujiepan/qwen3.8-moe-tiny-random finetune — 12.9M S — yujiepan 0 58
Vili-Elvis01/Qwen3.8-Max finetune — — — other Vili-Elvis01 1 0
inference-optimization/Qwen3.8-6.2B-A6.0B-small finetune — 6.24B M mit inference-optimization 0 48
Suhyoduh/Qwen3.8-2.4T-A95B-FP8 quantized -FP8 2.45T XXL other Suhyoduh 0 44
Blackfrost-Research/Qwen3.8-2.4T-A95B-DERISKED-BF16 finetune -BF16 2.45T XXL other Blackfrost-Research 0 14

Repository changelog

7 commits

Repository changelog 7 commits

DateCommitSHA
2026-08-12 Update README and add LICENSE 207bd68
2026-08-12 Update README and add LICENSE 1f1e318
2026-08-12 Update config.json b15b863
2026-08-11 Update model card bee0920
2026-08-11 Upload model card fa6bea3
2026-08-10 Update chat template 6c99963
2026-08-09 Upload folder using huggingface_hub f84a3db

All providers

7 providers · 7 endpoints

Where to run 7 providers · 7 endpoints

Hosted endpoints observed for Qwen3.8-2.4T-A95B, 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: $2.00/M.

ProviderModel refContextQuant$/M in$/M outtok/sTTFTToolsStatusObserved
Novita
hf, openrouter
 qwen/qwen3.8-2.4t-a95b 977K unknown $2.00 $6.00 32.9 1880 ms yes live 2026-10-25
Together AI
hf, openrouter
 Qwen/Qwen3.8-2.4T-A95B 986K unknown $2.00 $6.00 118.2 438 ms yes live 2026-10-25
DeepInfra
hf
 Qwen/Qwen3.8-2.4T-A95B 256K — $2.00 $6.00 72.7 284 ms yes live 2026-10-25
Alibaba
openrouter
 qwen/qwen3.8-2.4t-a95b 977K unknown $2.00 $6.00 — — yes live 2026-10-24
SiliconFlow
openrouter
 qwen/qwen3.8-2.4t-a95b 1M fp8 $2.00 $6.00 — — yes live 2026-10-24
Venice
openrouter
 qwen/qwen3.8-2.4t-a95b 256K unknown $2.00 $6.00 — — yes live 2026-10-24
Fireworks AI
hf
 accounts/fireworks/models/qwen3p8-2p4t-a95b 256K — — — 114.2 624 ms yes live 2026-10-25

Articles and coverage

31 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency.

Analysis9

Qwen/Qwen3.8-2.4T-A95B
modelscope.cn2026-08-08HN 31 pts · 4 comments
Qwen3.8 2.4T A95B: Intelligence, Performance and Price Analysis
artificialanalysis.ai2026-08-14authorityHN 5 pts · 0 comments
I Rented a 96 GB GPU and Took Uncensored Qwen3.8 From 44 to 125 tok/s
aseemshrey.com2026-08-24about the Qwen3.8 familyHN 7 pts · 3 comments
Serve Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72 | NVIDIA Technical Blog
developer.nvidia.com2026-08-12authority
Qwen3.8 2.4T A95B vs gpt-oss-120b (High)
artificialanalysis.aiauthority
What Is Qwen3.8-2.4T-A95B? Alibaba's 2.4T MoE
kie.ai2026-08-17
Qwen3.8-2.4T-A95B · LLM Model Review | CrucibleMark
cruciblemark.com2026-08-15
Qwen3.8 2.4T A95B in-depth review: 2.4 trillion parameter open source weight, a milestone for Alibaba Tongyi Qianwen – AI Dash
aidashxp.com2026-08-15
Qwen3.8-2.4T-A95B Benchmarks vs Opus 4.8 and GPT-5.6 Sol | MindStudio
mindstudio.ai2026-08-13

Community1

Qwen3.8 Fetches Weird URLs
news.ycombinator.com2026-08-20about the Qwen3.8 familyHN 8 pts · 0 comments
All coverage tracked for this model (31) — append-only, nothing is ever removed
TitleSourcePublishedFirst seenSignals
Qwen3.8 2.4T A95B vs gpt-oss-120b (High)artificialanalysis.ai—2026-10-07exa
Qwen3.8-2.4T-A95B: Specs, Benchmarks & Pricing | OpenLLMStackopenllmstack.com—2026-10-07exa
Qwen 3.8-2.4t-a95b Specs, Benchmarks & Token Pricing — Alibaba | ARMES Docsarmes.ai—2026-10-07exa
Qwen3.8-2.4T-A95B: release, architecture & GPU deployment | AIInfraInterviewsaiinfrainterviews.com—2026-10-07exa
Run QWEN3.8 27B on 16gb Nvidia GPUsgithub.com2026-09-172026-10-07hn
Qwen3.8-Maxqwencloud.com2026-09-022026-10-07hn
Show HN: Running 104GB Qwen3.8-Flash-Next on 48GB Mac with at ~12 tok/sgithub.com2026-09-012026-10-07hn
Run Qwen3.8 27B locally: real numbers from my Mac Studioterminalbytes.com2026-08-282026-10-07hn
Qwen3.8-Flash-Next Intelligence, Performance and Price Analysisartificialanalysis.ai2026-08-272026-10-07hn
Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapsesquesma.com2026-08-262026-10-07hn
Qwen3.8-Flash-Next Technical Report [pdf]github.com2026-08-262026-10-07hn
Qwen3.8-Flash-Nextqwen.ai official2026-08-262026-10-07hn
Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiencyqwen.ai official2026-08-262026-10-07hn
I Rented a 96 GB GPU and Took Uncensored Qwen3.8 From 44 to 125 tok/saseemshrey.com2026-08-242026-10-07hn
I found why Qwen3.8-27B was 14× slower on my 4070 Supergithub.com2026-08-232026-10-07hn
Qwen3.8 Fetches Weird URLsnews.ycombinator.com2026-08-202026-10-07hn
Qwen3.8-27B make medium the default effort level instead of xhighgithub.com2026-08-192026-10-07hn
Qwen3.8 27B scores 52 on Artificial Analysisartificialanalysis.ai2026-08-172026-10-07hn
Qwen3.8 27B at 256K: 50 TPS on a 24 GB GPUpiszczek.pl2026-08-172026-10-07hn
What Is Qwen3.8-2.4T-A95B? Alibaba's 2.4T MoEkie.ai2026-08-172026-10-07exa
Qwen3.8-2.4T-A95B · LLM Model Review | CrucibleMarkcruciblemark.com2026-08-152026-10-07exa
Qwen3.8 2.4T A95B in-depth review: 2.4 trillion parameter open source weight, a milestone for Alibaba Tongyi Qianwen – AI Dashaidashxp.com2026-08-152026-10-07exa
Qwen3.8-27Btwitter.com2026-08-142026-10-07hn
Qwen3.8 2.4T A95B: Intelligence, Performance and Price Analysisartificialanalysis.ai2026-08-142026-10-07hn
Qwen3.8-2.4T-A95B Benchmarks vs Opus 4.8 and GPT-5.6 Sol | MindStudiomindstudio.ai2026-08-132026-10-07exa
Serve Qwen3.8-2.4T-A95B, a 2.4T-Parameter Model, with Configurable Reasoning on NVIDIA GB300 NVL72 | NVIDIA Technical Blogdeveloper.nvidia.com2026-08-122026-10-07exa
Qwen3.8-2.4T-A95B: Alibaba's Open-Weight Qwen-Max Flagship Explained | MindStudiomindstudio.ai2026-08-122026-10-07exa
Qwen/Qwen3.8-2.4T-A95Bmodelscope.cn2026-08-082026-10-07hn
Qwen3.8 Max now ranked as the best overall model by agentic indexartificialanalysis.ai2026-08-062026-10-07hn
Qwen3.8-Max: A New Bar for Coding and Coworkqwen.ai official2026-08-032026-10-07hn
Qwen3.8-Maxtwitter.com2026-08-032026-10-07hn

Similar models

3 models

Similar models

  • DeepSeek-V4-Pro · 1.60T
    same size band (XXL), also MoE, same task (text-generation), permissive license
  • DeepSeek-V4-Pro-0813 · 1.65T
    same size band (XXL), also MoE, same task (text-generation), more providers, permissive license
  • Kimi-K2-Instruct · 1.03T
    same size band (XXL), also MoE, same task (text-generation), fewer providers

Sources

4 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:34 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 Qwen3.8-2.4T-A95B

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.