Unverified50% confidenceBenchmarkExact time
When the context exceeds 64,000 tokens, Strata keeps most of the KV cache in memory and retains only the hottest 32,000 tokens in VRAM, improving speed from approximately 51 to 63 tokens/second at a 262,000-token context.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/5/2026
First Seen
Valid until: 1/3/2027
Sources
Related Entities
Related Claims
UnverifiedDuring the prefill phase, Strata splits the prompt into chunks of up to 8,000 tokens and streams them over PCIe hidden behind computation, boosting the speed of processing a 32,000-token prompt from approximately 570 tokens/second to nearly 1,300 tokens/second.80% similarUnverifiedStrata uses low-bit quantization; the minimum configuration requires approximately 38 GB of combined memory and VRAM, the highest-quality configuration requires approximately 55 GB, and 64 GB of system memory is the ideal target.72% similarUnverifiedReal-world testing with the 284-billion-parameter DeepSeek V4 Flash shows token generation rising from 37 to 53 tokens/s (approximately 1.5x), and Prompt Processing jumping from 483 to 1,485 tokens/s (approximately three times).70% similarUnverifiedStrata converts the entire computer into a three-tier memory hierarchy: VRAM holds what each token touches, system memory pins all approximately 24,000 groups of experts, and the SSD stores a 29 GB N-gram lookup table.69% similarUnverifiedWhen running 4-bit DeepSeek R1 on the M3 Ultra, a 1,000-token context generates approximately 19 tokens per second using 392GB, dropping to approximately 15 tokens per second using 410GB at a 16,000-token context.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/978618API
curl https://kongchang.com/api/v1/knowledge/claims/978618MCP
get_claim(id=978618)