Unverified50% confidenceFactExact time
On an AMD Radeon RX 9060 XT 16GB graphics card using the IQ3 quantized version (10.9GB) and the llama.cpp Vulkan backend, the prompt itself had 104,813 tokens, 42% more than Astra processed for the same text; reading the hundred-thousand-plus tokens took nearly 30 minutes, and the model thought for 18 minutes and wrote 4,700 tokens without yet naming the game.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
11/4/2026
First Seen
Valid until: 11/18/2026
Sources
Can Qwen 3.8 27B Really Read Neural Network Weights? A 16GB Hands-On Test
youtubeOpaya11/4/2026
Related Entities
Related Claims
UnverifiedIn benchmark testing on an AMD RX 7900 GPU, Ornith 35B achieved an overall accuracy of 85% (18 out of 21 passed), a generation speed of 74 tokens/s, a prompt processing speed of 850 tokens/s, and a VRAM usage of 20 GB.78% similarUnverifiedUsing an AMD Radeon RX 7900XTX on Windows 11 to run Qwen3 via Strata, the generation speed (output tokens) is stable at 60–80 t/s, averaging approximately 68 tokens/s.69% similarUnverifiedOn an RTX 5070 with a 6-core Ryzen and 64 GB of memory, Strata achieves approximately 90 tokens/second for short conversations, approximately 67 tokens/second at a 128,000-token context, and approximately 1,300 tokens/second for reading prompts.69% similarUnverifiedOn 6 RTX 5090 GPUs, Colibri's measured speed reaches 9.0–9.2 tokens/s, while on a 25GB development machine streaming from NVMe the speed is 0.05–0.1 tokens/s.68% similarUnverifiedGFX1151 Engine's decoding (conversational generation) speed on the AMD 395 is 30 to 40 tokens per second, depending on the acceptance rate of speculative sampling.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1135727API
curl https://kongchang.com/api/v1/knowledge/claims/1135727MCP
get_claim(id=1135727)