Toolbit.aiToolbit.ai

Find, compare, and explore the best AI tools to match your specific tasks and use cases.

Explore

  • AI Search
  • Compare ToolsNew
  • Browse Categories
  • Trending Tools
  • Most Popular
  • New Additions

Resources

  • Updates HubNew
  • AI News
  • ModelsNew
  • Blog Articles
  • NewsletterNew

Company

  • Launch a Tool
  • Advertise with Us
  • Guest Post
  • Contact Us
© 2026 Toolbit.ai. All rights reserved.
Privacy PolicyTerms & ConditionsDisclaimer
There's An AI For That favicon
There's An AI For That•The front page of AI for everyone
Toolbit.ai
Toolbit.ai
UpdatesNew
Blog
Toolbit.ai
Toolbit.ai
Toolbit.ai
Toolbit.ai
UpdatesNew
Blog
Sign in
  1. Home
  2. Updates
  3. Models
  4. JT-4.1 Flash 236B A21B
CM
China Mobile
Released July 9, 2026

JT-4.1 Flash 236B A21B

JT-4.1 Flash 236B A21B by China Mobile: a 236B/21B MoE non-reasoning model with abstention-based hallucination control.

Announcement
Inputs
Text
Outputs
Text

Model Overview

Capabilities, design details, and architectural traits

JT-4.1 Flash 236B A21B - China Mobile's non-reasoning MoE model with abstention-based hallucination control

JT-4.1 Flash 236B A21B is a Mixture of Experts model from China Mobile with 236B total parameters and 21B activated parameters. It is explicitly a non-reasoning architecture model.

TraitDetail
Architecture236B total, 21B activated (MoE), non-reasoning
Hallucination controlReduced hallucination rate by 20 points through increased abstention rate

Abstention-driven reliability

A defining characteristic of JT-4.1 Flash is its approach to reducing hallucination by raising its abstention rate - the model declines to answer when uncertain rather than generating potentially incorrect responses. This strategy cut its hallucination rate by 20 points, demonstrating that a non-reasoning architecture can still achieve competitive reliability through selective answering.

Benchmark Performance

Independent evaluations · Artificial Analysis

39.9%
Intelligence
52.4%
Coding Index
34.9%
Agentic Index

Accuracy & Capability Details

GPQA - Graduate Science84.5%
Humanity's Last Exam17.8%
SciCode - Scientific Coding38.2%
Long Context Reasoning70.3%
Specs
Context window
-tokens

Prices in USD.

Compare Models Side-by-Side

Evaluate specifications, pricing, and independent benchmark indices

Model Details
General Info
ProviderChina MobileAnthropicAnthropic
Release DateJuly 9, 2026July 24, 2026June 9, 2026
Knowledge Cutoff-May 2026-
Context & Limits
Context Window-1M1M
Pricing (per 1M tokens)
Input Pricing
Free
Best Input Pricing
$5$10
Output Pricing
Free
Best Output Pricing
$25$50
Modalities
Inputs
text
textimage
textimagefile
Outputs
text
text
text
Benchmarks (0-100)
Intelligence Index39.9
63.1
Best Intelligence Index
62.1
Coding Index52.4
78.0
Best Coding Index
76.5
Agentic Index34.9
59.2
Best Agentic Index
56.6
JT-4.1 Flash 236B A21B
Claude Opus 5
Claude Fable 5

GPQA Benchmark

Graduate-level reasoning and expert Q&A evaluation.

85%
JT-4.1 Flash 236B A21B
GPQA Benchmark
Score: 85%
CM
JT-4.1 Flash 236B A21B
93%
Claude Opus 5
GPQA Benchmark
Score: 93%
Claude Opus 5
93%
Claude Fable 5
GPQA Benchmark
Score: 93%
Claude Fable 5

Humanity's Last Exam

Extremely difficult logical reasoning and knowledge.

18%
JT-4.1 Flash 236B A21B
Humanity's Last Exam
Score: 18%
CM
JT-4.1 Flash 236B A21B
55%
Claude Opus 5
Humanity's Last Exam
Score: 55%
Claude Opus 5
56%
Claude Fable 5
Humanity's Last Exam
Score: 56%
Claude Fable 5

Long Context Reasoning

Logical reasoning over long context windows.

70%
JT-4.1 Flash 236B A21B
Long Context Reasoning
Score: 70%
CM
JT-4.1 Flash 236B A21B
76%
Claude Opus 5
Long Context Reasoning
Score: 76%
Claude Opus 5
77%
Claude Fable 5
Long Context Reasoning
Score: 77%
Claude Fable 5

Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.

Back to all models