Sources
·
Twitter/X
@antirez
Active
antirez
Linked items
100
Errors
37
Last fetched
2026-05-14 00:44
Priority
P3
Recent items
The 15 most-recently-ingested items from this source, newest first.
| Title |
Type |
Fetched |
|
@antirez: About a recent article that had, IMHO, not very strong arguments (for me), but that was commented en...
|
article |
2026-05-13
|
|
@antirez: Not bad. Committed.
antirez (@antirez)
If the correctness tests will pass, expect a sensibly fa...
|
article |
2026-05-13
|
|
@antirez: If the correctness tests will pass, expect a sensibly faster DS4 inference speed for DGX Spark, and ...
|
article |
2026-05-13
|
|
@antirez: Maybe very little folks noticed that DS4 supports both generating the refusal vector (or whatever be...
|
article |
2026-05-12
|
|
@antirez: Announcing with gratitude that @audreyt just gifted me an M5 Max 128GB MacBook Pro! It will let me d...
|
article |
2026-05-12
|
|
@antirez: About M5 support, if you are at a company that is big enough to send me an M5 Max 128GB Mac, conside...
|
article |
2026-05-12
|
|
@antirez: Btw I'm truly honored for this contribution. Thanks @ggerganov. Make sure to read the DwarfStar4 REA...
|
article |
2026-05-12
|
|
@antirez: Breaking! @ggerganov himself confirming that DS4 2 bit quants actually work very well! He is running...
|
article |
2026-05-12
|
|
@antirez: The new Dwarf Star 4 weights download script features a new q4-imatrix GGUF now. I compared it again...
|
article |
2026-05-12
|
|
@antirez: When GPT 5.5 tells me to be over capacity I think at all the questionable automatic day and night au...
|
article |
2026-05-12
|
|
@antirez: I need to run more benchmarks but in another test I also found, as I expected, DS4 to be able to sol...
|
article |
2026-05-12
|
|
@antirez: Qwen 3.6 27B is a great model, but if you believe that an extremely good 284B parameters model is th...
|
article |
2026-05-12
|
|
@antirez: Useless to say this is even truer with standardized benchmarks where benchmaxing is a real temptatio...
|
article |
2026-05-12
|
|
@antirez: Moreover compared to Qwen the DS4F prose was always a lot better, and as always it thinks proportion...
|
article |
2026-05-12
|
|
@antirez: Btw, results in this "facts recall": obviously if you have 10x parameters it shows up. DeepSeek v4 F...
|
article |
2026-05-12
|