Issues

[Feature Request] Support JANGTQ / JANG Quantization format (e.g., Kimi-K2.7-Code-JANGTQ_K)

#2351

· super-lsp opened

on Jul 23, 2026

Feature request: Qwen3-Embedding (text) support with last-token/EOS pooling on /v1/embeddings

#2350

· youngjin39 opened

on Jul 23, 2026

Severe prefill performance regression on Jundot/Qwen3.6-35B-A3B-oQ4e-mtp

#2349

· AndrewY-unique opened

on Jul 23, 2026

Can't load quantized GLM-5.2-oQ3.5e-mtp after using the built-in oQ Quantization

#2347

· noonkho opened

on Jul 23, 2026

HF Model Downloads Fail

#2345

· carlyd95 opened

on Jul 23, 2026

SpecPrefill × TurboQuant KV: the BatchTurboQuantKVCache half of #766 still deadlocks on 0.5.3 (SpecPrefill cache never quantised + non-recovering cache-error path)

#2336

· necoli1822 opened

on Jul 22, 2026

Bonsai 1-bit prefill (M>5) falls back to fp16 dequant: ~16x memory blowup, ~150x slower

#2335

· duffercn opened

on Jul 22, 2026

Fatal teardown watchdog (60s) fires on normal unload of a 420GB engine, exits mid-teardown and strands ~381GB wired (reboot-only, #2184 end state)

#2334

· anicaise-ai opened

on Jul 22, 2026

Claude Code Cache Performance Issues

#2333

· JJOne123 opened

on Jul 22, 2026

Schema-blind per-parameter json.loads() fallback (3 branches: Qwen/Llama, GLM, MiniMax) silently drops array/object tool-call params to raw strings on minor JSON syntax errors

#2332

· bobacus opened

on Jul 22, 2026

Expose VLM processor kwargs through the OpenAI-compatible API for Unlimited-OCR multi-page mode

#2331

· BeCreatedP opened

on Jul 22, 2026

0.5.2: two concurrent identical fully-cached prefills wedge the scheduler; phantom waiting request survives disconnect; teardown then impossible (strands wired)

#2330

· anicaise-ai opened

on Jul 22, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

All issues

[Feature Request] Support JANGTQ / JANG Quantization format (e.g., Kimi-K2.7-Code-JANGTQ_K)

Feature request: Qwen3-Embedding (text) support with last-token/EOS pooling on /v1/embeddings

Severe prefill performance regression on Jundot/Qwen3.6-35B-A3B-oQ4e-mtp

Can't load quantized GLM-5.2-oQ3.5e-mtp after using the built-in oQ Quantization

HF Model Downloads Fail

SpecPrefill × TurboQuant KV: the BatchTurboQuantKVCache half of #766 still deadlocks on 0.5.3 (SpecPrefill cache never quantised + non-recovering cache-error path)

Bonsai 1-bit prefill (M>5) falls back to fp16 dequant: ~16x memory blowup, ~150x slower

Fatal teardown watchdog (60s) fires on normal unload of a 420GB engine, exits mid-teardown and strands ~381GB wired (reboot-only, #2184 end state)

Claude Code Cache Performance Issues

Schema-blind per-parameter json.loads() fallback (3 branches: Qwen/Llama, GLM, MiniMax) silently drops array/object tool-call params to raw strings on minor JSON syntax errors

Expose VLM processor kwargs through the OpenAI-compatible API for Unlimited-OCR multi-page mode

0.5.2: two concurrent identical fully-cached prefills wedge the scheduler; phantom waiting request survives disconnect; teardown then impossible (strands wired)

Search results