Ctrl K
Back to models

Alibaba

Qwen3.8 2.4T A95B

  • Released
  • Open weights
  • Qwen3.8-Max License
Compare
Intelligence Index

Reasoning setting

39.9#21 of 83measured models
Measured · Artificial Analysis v4.3.2
API price from
$2.00 in / $6.00 out / 1M tokOpenRouter · SiliconFlow · FP8
Context window
983.62K tok

Performance profile

ReasoningRank #6 of 107
93.5%
GPQA Diamond
CodingRank #22 of 94
54.1%
SciCode
Cost efficiencyRank #59 of 67
18.5pts / $
Intelligence ÷ eval cost

Capabilities

Context window

Modalities

Measured setting

Integration

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
Qwen3.8 2.4T A95B
Cost per task (USD)
Qwen3.8 2.4T A95BReasoning
Intelligence Index
39.9
Cost / task
$2.16
Evaluation time
22.6m

Pricing & access

API route
OpenRouter · SiliconFlow · FP8API
Tier & contextStandard

USD per 1M tokens

Input
$2.00
Output
$6.00
Cached input
$0.25

Run locally

GGUFQ8_0
Context
Estimated memory

~2,665.1 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

2816 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
2,421.7 GiB
Context + state
1.3 GiB16-bit cache
App allowance
242.2 GiB
Download
2,600.2 GB56 files · keep together
Disk space
~2866 GB
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · Q8_0 · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/unsloth/Qwen3.8-2.4T-A95B-GGUF:Q8_0
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter unsloth/Qwen3.8-2.4T-A95B-GGUF, choose Q8_0, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

vLLM setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

Qwen3.8 2.4T A95B · Reasoning#22 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 19Qwen3.8 Max (0902)Reasoning43.1%
  6. 20GPT-5.6 TerraMax42.9%
  7. 21GPT-5.3-CodexExtra high42.5%
  8. 22Qwen3.8 2.4T A95BReasoning42.4%
  9. 23GLM-5.3Max42.3%
  10. 24DeepSeek-V4-Pro-0813Max41.0%
  11. 25GLM-5.3-FlashMax39.9%

Peers use their best measured result on this benchmark.

Alternatives

Each model at its highest-scoring measured setting · Intelligence Index v4.3.2 and its evaluation cost per task.