JT-4.1 Flash 236B A21B
JT-4.1 Flash 236B A21B by China Mobile: a non-reasoning Mixture of Experts model that lowers hallucination through abstention-based answering when unsure.
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.
| Trait | Detail |
|---|---|
| Architecture | 236B total, 21B activated (MoE), non-reasoning |
| Hallucination control | Reduced 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
Accuracy & Capability Details
Compare Models Side-by-Side
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | China Mobile | Anthropic | Anthropic |
| Release Date | July 9, 2026 | September 22, 2026 | September 28, 2026 |
| Knowledge Cutoff | - | - | Jun 2026 |
| Context & Limits | |||
| Context Window | 262K | 1M Best Context Window | 1M Best Context Window |
| Pricing (per 1M tokens) | |||
| Input Pricing | Free Best Input Pricing | $4 | $2 |
| Output Pricing | Free Best Output Pricing | $20 | $10 |
| Modalities | |||
| Inputs | text | textimagefile | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 27.3 | 57.6 Best Intelligence Index | 56.0 |
| Coding Index | 52.4 | - | - |
| Agentic Index | 34.9 | - | - |
Humanity's Last Exam
Extremely difficult logical reasoning and knowledge.
Long Context Reasoning
Logical reasoning over long context windows.
SciCode Benchmark
Scientific coding and mathematical modeling.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
Top AI Models
Leading alternatives by intelligence score
