pigeon57434

• LM Studio finally added support for MTP Speculative Decoding

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 99 comments

• People seem to already not care about heretic?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 35 comments

• What happened to Black Forest Labs?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 53 comments

• I know this community doesn't like AA-II much but this is a NUTTY score for Kimi-K2.6 wow

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 1 comments

• Heretic has FINALLY defeated GPT-OSS with a new experimental decensoring method called ARA

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 152 comments

• Qwen3.5-27B as good as DeepSeek-V3.2 on AA-II (plus some more data)

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 28 comments

• Averaged over the 36 text benchmarks provided for Qwen3.5's new small models I have a question

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 8 comments

• What's the best open source speech to text model

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 17 comments

• why do uncensored and abliterated fine tunes still refuse my request

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 49 comments

• Qwen/Qwen3-4B-Thinking-2507

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 6 comments

• How do you change the system prompt in the oobabooga text gen webui?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 5 comments

• DeepSeek-R1-0528-Qwen3-8B optimal settings?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 6 comments

• How much does flash attention affect intelligence in reasoning models like QwQ

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 21 comments

• QwQ-32B seems to get the same quality final answer as R1 while reasoning much more concisely and efficiently

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 129 comments

• can someone explain all the different quant methods

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 30 comments

• Qwen team seems sure that their model is better than LiveBench ranks it and demand a rerun with more optimal settings, which is crazy because it already performed really great

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 28 comments

• fal announces Flux a new AI image model they claim its reminiscent of Midjourney and its 12B params open weights

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 120 comments

• Elon says Grok 2 won't be open sourced until Grok 3 is stable which isn't for a couple MONTHS

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 11 comments

• Which do you think will be better: Qwen-3 or Llama-4

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 33 comments

• I calculated the effective cost of R1 Vs o1 and here's what I found

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 18 comments

• open source model small enough to run on a single 3090 performing WAY better in most benchmarks than the ultra proprietary closed source state of the art model from only a couple months ago

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 22 comments

• Whats the current State-of-The-Art for voice cloning?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 4 comments

• Now that Phi-4 has been out for a while what do you think?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 70 comments

• local solutions for government?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 23 comments

• I've been out of the local llm space for a while what do people use to run these models now?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 21 comments

• Has anyone tried Dolphin-Vision yet?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 51 comments

• When do you think 1-bit LLMs will actually kick off if ever?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 73 comments

• What's the best model I can run with my GPU?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 22 comments

• Gemma2-27B outperforms both Qwe2-72B and Llama3-70B on LMSYS. But is it actually better?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 111 comments

• SD3-Medium is on imgsys rankings now it shows 62 ELO better than SDXL base

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 46 comments

• SD3 support in A1111?

  Posted by pigeon57434@reddit | LocalLLaMA | View on Reddit | 7 comments