Models & Labs
Anthropic Holds Back Mythos Model
Lev Selectormedium confidence
Why it matters
- →The delay underscores the importance of safety in AI model deployment, impacting future releases.
Anthropic has decided to hold back the release of its highly anticipated Mythos model due to safety concerns. The company is prioritizing the responsible deployment of AI technologies and is taking additional time to ensure that the model meets safety and ethical standards. This decision highlights the ongoing challenges in balancing innovation with safety in AI development.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1Anthropic Faces Pressure on AI Weapons Development
AI Explained · February 27, 2026 · Background
- 2Anthropic unveils powerful AI through Project Glasswing
The Rundown AI · April 8, 2026 · Background
- 3Anthropic Previews Claude Mythos AI System
The AI Advantage · April 8, 2026 · Related
- 4Claude Mythos Model Released with 244-Page Report
AI Explained · April 8, 2026 · Background
What happened next
- 5Anthropic Launches Fable 5 Model
Sam Witteveen · June 9, 2026 · Related
- 6US Halts Anthropic's Mythos 5 AI Model Release
The Verge AI · June 16, 2026 · Related
- 7Anthropic Faces U.S. Government Standoff Over Mythos
The Rundown AI · June 18, 2026 · Related
- 8Anthropic's Mythos 5 Returns for Select Users
The Verge AI · June 27, 2026 · Related
- 9Commerce Department Limits Access to Anthropic's Mythos
The AI Daily Brief · June 30, 2026 · Related
Follow this story
Open the full story →Anthropic Holds Back Mythos Model
3 developments
- Apr 10 · Lev Selector
Anthropic Holds Back Mythos Model (This article) - Apr 10 · Fireship
Anthropic restricts access to Mythos model↳ Anthropic cites safety concerns as the reason for restricting general user access to the Mythos model. - Apr 23 · The Rundown AI
Anthropic's Mythos AI model reportedly leaked↳ The leaked Mythos model was accessed by a Discord group using stolen credentials.
More in Models & Labs
Models & Labsother
llama.cpp b12340 adds CUDA 13 and ROCm 10.0
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13 and ROCm 10.0 across Linux and Windows. Users with newer NVIDIA GPUs or AMD accelerators no longer need to compile from source to get performance, while Apple Silicon builds now ship KleidiAI disabled by default, shifting the focus back to standard ARM optimizations. The inclusion of Snapdragon support for Linux and Android marks a significant step toward on-device AI beyond the x86/ARM desktop ecosystem. It is less about new features and more about ensuring llama.cpp remains the universal runtime for every available accelerator.
llama.cpp Releases
Models & Labscoding
llama.cpp b12342 adds few-row MMQ for speculative decoding
Speculative decoding just got a significant speed boost on NVIDIA GPUs. This release introduces a specialized kernel for 'few-row' matrix multiplications, targeting the small batch sizes typical of verification passes in speculative decoding workflows. By optimizing these specific operations for quantized formats like Q8_0 and IQ4 variants, llama.cpp reduces latency where it matters most during token generation. The change is CUDA-only for now but leverages hardware capabilities from Ampere onwards, making local inference faster without requiring new model architectures or complex configuration changes.
llama.cpp Releases
© TechCrunch AIModels & Labsmodels
Google deploys Gemini 4 Hydrogen for faster Search
Google is quietly shifting its search infrastructure by replacing older fast models with Gemini 4 Hydrogen, a tier designed specifically for speed and cost-efficiency rather than raw reasoning power. This model completes tasks 2.5 times faster than its predecessor while beating the more powerful Argon on specific intelligence benchmarks, proving that specialized efficiency can outperform brute force in high-volume workflows. By integrating this into Search's fan-out technique, Google now breaks down complex queries into subqueries with greater precision and speed, delivering richer visual responses without the latency of heavier models. This marks a strategic pivot toward agentic execution layers within consumer search, making AI Mode feel snappier for everyday multi-part questions.
TechCrunch AI