NOUMENOpen app
NOUMEN

Meta · Open weights

Sao10K Stheno 8b

Added Jan 1, 2024Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Sao10K's latest Stheno fine-tune optimized for instruction following.

Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.20
Output
$0.20

Reference prices as published by the model provider. API example

Specifications

Context window
16.4K
Max output
8.2K
Average output (7 days)
177 tokens
Longer than 18% of 468 models

Side by side

Sao10K Stheno 8b compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Sao10K Stheno 8bMetaThis modelMuse Spark 1.3MetaRemove
ProviderMetaMeta
Discovery tierOpen weightsFrontier
AddedJan 1, 2024Sep 2, 2026
Limits
Context window16.38K1.05MBest
Max output8.19K—
Average output177 tokens1.2K tokens
Pricing, USD per 1M tokens
Input$0.20Best$1.25
Output$0.20Best$4.25
Cache read—$0.15
Performance
Throughput30.4 tok/s274.4 tok/sBest
Time to first token2.1 sBest2.9 s
Request success (24h)67.2%100.0%Best
Benchmark indexes
Intelligence Index—48.1
Coding Index—75.8
Capabilities
ReasoningNoYes
ToolsNoYes
VisionNoYes
Structured outputNoYes
PDF inputNoYes
Audio inputNoYes
Video inputNoYes
Reasoning effort—minimal, low, medium, high, xhigh

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Sao10K Stheno 8b
  • Muse Spark 1.3

Intelligence Index

  • Sao10K Stheno 8b—
  • Muse Spark 1.348.1
3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 75.8

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 57.9%

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 29.8%

  • Harvey LAB-AAv1.1

    Legal tasks passed in full, with no material hallucination

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 8.9%

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 1,581 Elo

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 1,681 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 26.6%

  • AA-LCRv1.1

    Reasoning across long documents

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 83.0%

  • MLCR-AA

    Long-context reasoning over medical records

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 13.9%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 48.7%

  • GPQA Diamond

    Graduate-level science questions

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 93.5%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 33.3%

  • SciCode

    Scientific computing in Python

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 58.8%

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Sao10K Stheno 8b: —
    • Muse Spark 1.3: 83.0%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
30.4 tok/s
Faster than 21% of 254 models
Time to first token
2.1 s
Quicker than 74% of 254 models
Prompt cache hits
0.0%
Request success (24h)
67.2%

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Sao10K/L3-8B-Stheno-v3.2",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="Sao10K/L3-8B-Stheno-v3.2",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Meta

Other Meta models, starting with the same family.