DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows....
Tested on 15 benchmarks · BenchGecko score 60.1. Top scores: OTIS Mock AIME 2024-2025 (94.4%), ARC-AGI (89.0%), GPQA diamond (88.0%).
Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.
Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.
ARC-AGI 2, harder sequel to ARC. More complex abstract reasoning patterns that test generalization ability beyond training data.
Deceptively simple questions that humans find easy but AI models often get wrong. Tests common sense and reasoning gaps.
Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.
- Typetext
- Context1.0M tokens (~524 books)
- ReleasedJul 2026
- LicenseProprietary
- StatusActive
- Cost / Message~$0.001
Frequently Asked Questions
Key facts · as of 2026-10-05
- DeepSeek V4 Flash 0731 by DeepSeek. BenchGecko score 60.1, rank 86 of 312 scored models (normalized average of public benchmark scores).
- List price $0.0152 input · $1.28 output per 1M tokens (as of 2026-10-05).
- Sold by 27 providers (as of 2026-10-05): Relace (fp4) $0.0152 in / $1.28 out · Sail Research (fp4) $0.0190 in / $0.30 out · Sail Research (fp4) $0.0190 in / $0.42 out · OpenInference (fp4) $0.0198 in / $1.66 out · Reka $0.0210 in / $0.53 out · and 22 more. Every provider
- Gecko Tests: Who Are You E (Sometimes says it is Google or Anthropic) · Tokenizer Tax D (68% more tokens outside English).
How to cite · data as of 2026-10-05
DeepSeek V4 Flash 0731 · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/deepseek-v4-flash-0731
Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP