Ds4 v0.6.2: Runs DeepSeek 284B on single DGX Spark at 1000 tok/s
pbaylies · x · 2026-08-19
Ds4 v0.6.2 introduces a real memory budgeting mechanism that measures actual usage instead of static reservation, enabling robust scaling in shared environments. Benchmarks show it can stably serve the 284B parameter DeepSeek V4 Flash model on a single DGX Spark at 1000 tok/s prefill and 59 tok/s for multi-agent serving. The design implements demand-mapped context and graceful rejection upon resource exhaustion instead of OOM crashes.
More from Infra
- Prediction: Opus 5.5-level models on laptops by 2027, backed by Qwen 27B data — cephaloform · 2026-10-16
- Oracle down 12% in a week, nearly 50% off June high as OpenAI exposure spooks market — alvelda · 2026-10-16
- US data center moratoriums top 400 — an AI bull explains why they're actually good for AI — kevinsxu · 2026-10-16
- Microsoft Research unveils Magnetar, the fastest zkVM at 10-39x speedups and post-quantum security — jedisct1 · 2026-10-16
- Bonsai 2 2B hits 120 tok/s on iPhone 18 Pro, outpaces Qwen 3.5 2B at half the memory — cephaloform · 2026-10-16
- Samsung cuts Q4 phone output as Nittobo shifts E-glass to AI-chip T-glass — zephyr_z9 · 2026-10-16