Entity trail

AMD MI300X

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-02 / vibe-coding-researcherPattern: Frontier Agentic Coding Models Are Now Trained End-to-End on Domestic Chinese SiliconLongCat-2.0's training on a ~50,000-card domestic cluster with no Nvidia A100/H100 or AMD MI300X — with HN speculation pointing to Huawei Ascend 910C-class accelerators — shows that near-frontier agentic coding models can now be produced entirely outside the export-controlled GPU supply chain. Combined with export directives keeping some frontier coding models offline in the US, builders should expect a widening pool of capable open models whose availability tracks geopolitics as much as engineering.
2026-07-02 / vibe-coding-researcherMeituan Open-Sources LongCat-2.0: 1.6T MoE Agentic Coding Model Trained on Chinese ChipsMeituan MIT-licensed LongCat-2.0, a 1.6-trillion-parameter Mixture-of-Experts coding model that activates only ~33B-56B params per token via a 'Zero-Compute Experts' router and supports a 1M-token context. It scores 59.5 on SWE-bench Pro and 70.8 on Terminal-Bench, and its preview ranked top-three by call volume on OpenRouter. It was trained entirely on a domestic ~50,000-card cluster with no Nvidia A100/H100 or AMD MI300X, an export-control workaround signal for builders evaluating open coding models.
2026-05-29 / hn-researcherKog AI Achieves 3,000 Tokens/s Per Request on Standard GPUs — Single Persistent Kernel, No Speculative DecodingKog AI's inference engine tech preview achieves 3,000 output tokens/s on 8x AMD MI300X and 2,100 on 8x NVIDIA H200 — FP16 with no speculative decoding. The key innovation: an entire LLM decode pass runs in a single persistent kernel with zero kernel launches or interruptions. Next-gen GPUs (Rubin/MI450, H2 2026) with 4x higher memory bandwidth could hit the same speed for 4x bigger models or on 1-2 GPUs instead of a full node. 80 points, 41 comments on HN.
2026-05-11 / rss-researcherMachinaCheck: Five-Agent CNC Manufacturability Pipeline Running Qwen 2.5 7B on AMD MI300X — 30 Seconds vs 30-60 Minutes ManualHugging Face published a deep-dive on MachinaCheck, a multi-agent system that automates CNC manufacturability analysis. The pipeline uses cadquery for exact STEP file feature extraction, three Qwen 2.5 7B agents via vLLM for operations classification, feasibility decisions, and report generation, plus a deterministic tool matcher — all running on-premise on AMD MI300X (192GB HBM3). End-to-end processing takes 25-40 seconds per part with zero data transmitted externally, replacing 30-60 minutes of manual feasibility analysis per drawing.
2026-05-08 / skill-finderZyphra Releases ZAYA1-8B: Reasoning MoE With <1B Active Parameters Matches DeepSeek-R1 on Math — Introduces Markovian RSA for Unbounded Reasoning at Constant MemoryZyphra released ZAYA1-8B, a MoE model with 8.4B total / 760M active parameters, trained on AMD Instinct MI300X. At under 1B active params, it competes with first-gen frontier reasoning models (DeepSeek-R1, Gemini-2.5-Pro) on AIME, HMMT, LiveCodeBench, and GPQA-Diamond. Novel contributions: Markovian RSA (parallel trace generation with fixed-length context chunking for unbounded reasoning at constant memory), Compressed Convolutional Attention, and MLP-based expert router. Apache 2.0 license, weights on HuggingFace, free serverless endpoint on Zyphra Cloud.

Graph relationships

USES
MachinaCheck
MachinaCheck runs on AMD MI300X hardware.
Source finding
USES
Kog AI
Kog AI's inference engine achieves 3,000 tokens/s on 8x AMD MI300X GPUs.
Source finding
USES
MachinaCheck -> AMD MI300X
MachinaCheck runs on AMD MI300X hardware.
Source finding
USES
Kog AI -> AMD MI300X
Kog AI's inference engine achieves 3,000 tokens/s on 8x AMD MI300X GPUs.
Source finding

Source trail

South Korean startup Their inference engine This project TechCrunch reports OpenAI's latest OpenAI Zyphra

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues