David Ha runs K3 locally on M5 Max

// 1h agoNEWS

David Ha runs K3 locally on M5 Max

AI researcher David Ha (@hardmaru) shared an experiment running the massive K3 language model locally on Apple's M5 Max chip. Operating at a throughput of approximately 0.3 tokens per second, the test demonstrates the capability of high-capacity Apple Silicon unified memory to host huge models, even if the current performance is exceptionally slow.

// ANALYSIS

Local execution of massive frontier models on workstation hardware is technically impressive but currently constrained by severe compute and memory bandwidth bottlenecks.

–Unified memory architecture allows consumer-tier workstations to fit multi-hundred-billion parameter models that previously required enterprise GPU clusters.
–An inference speed of 0.3 tokens per second is impractical for interactive workflows, serving primarily as an early proof-of-concept.
–Future optimizations in quantization, MoE sparse execution, and specialized matrix kernels will likely improve local token generation speeds substantially.

// TAGS

k3local-aiapple-siliconm5-maxllm-inferencehardware

DISCOVERED

1h ago

2026-07-28

PUBLISHED

1h ago

2026-07-28

RELEVANCE

6/ 10

AUTHOR

hardmaru

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE1h ago

Mintlify editor loads pages 9.5x faster

Mintlify has introduced a major performance update to its documentation platform, enabling its editor to load pages 9.5x faster than before. The enhancement aims to reduce latency and friction for developers writing, editing, and managing technical documentation and API reference pages.

OPEN SOURCE2h ago

OpenAI open-sources Codex Security CLI

OpenAI has open-sourced Codex Security CLI, a tool designed to help developers discover, validate, and remediate vulnerabilities directly within their local development environment and CI/CD pipelines. By combining local context analysis with AI threat modeling, Codex Security CLI allows engineering teams to scan local repositories, automate security checks in workflows like GitHub Actions, and track security findings over time.

UPDATE2h ago

BAI adds four frontier models to API gateway

BAI, an API platform offering unified access to large language models, has expanded its offerings with four new frontier AI models. The update aims to empower developers by shortening the time required to deploy cutting-edge AI capabilities into applications through a centralized gateway.