CAISI releases evaluation of DeepSeek AI models
National Institute of Standards and Technology's Center for AI Standards and Innovation (CAISI) has released an official evaluation of multiple artificial intelligence models from China-based DeepSeek. This review provides an authoritative U.S. government technical analysis regarding the operational reliability and risk profiles of DeepSeek's widely used model portfolio. Claims are as reported; this summary makes no determination about accuracy or significance.
- First detected
- Aug 25, 2026
- Last updated
- Aug 27, 2026
Low confidence
Based on available reporting; the relationship between the publisher and the event is unclear.
Unconfirmed
No independent reporting recorded yet.
What does this mean?
Corroboration measures how many genuinely independent sources support the event. Confidence measures how reliable the available evidence appears.
Stable
No recent reporting has materially changed the known facts.
Follow this development to see meaningful updates as new evidence emerges.
Save keeps this for later. Follow tracks meaningful changes as new evidence emerges — it shapes your Following Feed, alerts, and digest eligibility, and doesn't promise an instant notification.
What we know
Organizations & participants
- Organization: Center for AI Standards and Innovation (CAISI)
Product
- Product: DeepSeek models
Availability
- Availability: General availability
- Available in: United States
Why it matters
Widespread enterprise interest in DeepSeek models has surged due to cost efficiencies and competitive benchmark performance. NIST's risk findings establish an official baseline for enterprise risk managers, compliance teams, and security officers assessing foreign-developed open and proprietary models. Organizations deploying these models must review their risk posture against CAISI findings to ensure adequate safeguards and compliance with emerging standards.
Coverage
Primary/vendor sources vs independent reporting
Primary / vendor source: information published directly by the company, organization, government body or project involved. Useful as a primary source, but not independent confirmation.
Independent reporting: reporting or analysis from a source independent of the organization making the underlying claim.
How this developed
Aug 25, 2026
Development detected
Sep 30, 2025
New reporting added
Related Intelligence
- DevelopmentStableAlso involving China
Mistral unveils new AI model
Mistral has unveiled a new AI model that the company claims rivals leading open AI systems from China, according to CNBC. The release comes amid intensifying competition between Western and Chinese developers in the open-weight and open-source model space. Claims are as reported; this summary makes no determination about accuracy or significance.
3 independent sources - DevelopmentStableAlso involving China
Reflection AI launches Beam 501B open-source model
Reflection AI Inc. has launched Beam, an open-source large language model featuring 501 billion parameters. The release follows the startup reaching a $25 billion valuation and reportedly entering a $6.3 billion agreement with SpaceX to rent Nvidia GB300 NVL72 systems used for the initiative. Claims are as reported; this summary makes no determination about accuracy or significance.
2 independent sources - DevelopmentNewAlso involving China
DeepSeek collaborates with Huawei to design AI chips
DeepSeek is collaborating closely with Huawei Technologies to guide the architectural design of future AI semiconductors, according to reporting from The Information. Huawei engineers are reportedly incorporating DeepSeek's technical input on calculation and inter-processor data transfer into AI chips planned for release next year and beyond, aiming to support frontier AI development independently of Nvidia hardware. Claims are as reported; this summary makes no determination about accuracy or significance.
- DevelopmentNewAlso involving China
MiniMax released M3.1 Flash
MiniMax has released M3.1 Flash, scoring 42 on the Artificial Analysis Intelligence Index—a 13-point increase over its predecessor—at half the operational cost. The model also scored 73.7 percent on the DeepSWE v1.1 coding benchmark at under $0.30 per task, matching Deepseek V4.1 Flash 1017 across both benchmark performance and pricing metrics. Claims are as reported; this summary makes no determination about accuracy or significance.