Daily Hub scan · 90-day window

New open-weight models

Recently created text-generation and image-text-to-text releases from the public Hugging Face Hub API. Known labs plus trending rows. Finetune spam with low downloads stays off. GGUF packs are searched on unsloth, bartowski, lmstudio-community, and ggml-org. Q3–Q8 only — Q1/Q2 stay off. Fit is a labeled Q4_K_M estimate, not a measured card. Runs on is evidence-only: llama.cpp convert arch or a listed GGUF pack, vLLM’s supported-architectures list, Ollama library slugs, Strata’s Qwen 3.8 Flash notes on /backends, and SGLang architecture names when listed. LM Studio and Lemonade follow GGUF as llama.cpp wrappers. Empty stays unknown. MLX is off — not a Radeon path.

79 models · as of 2026-10-16. The table reads /data/new-models.json in the browser so a server cron can refresh it without a rebuild. This page ships a static fallback from the last sync. Feed: /new-models/rss.xml.

Open calculator House models Quants Can my Radeon run it? How we test

Filter

The last 90 days

79 rows

Newest release first, grouped by day. Click a column to sort inside each day. Select a row for the detail card. Without JavaScript the fallback table stays visible.

Size band

GGUF

Fits my GPU

Runs on

Estimate — not a measured card. Param counts come from Hub safetensors or config. Q4_K_M GiB is a Hub file size when a pack lists one, otherwise params × 4.85 bits. Fit adds 1.5 GiB overhead against 16 / 24 / 32 / 2×32 GB. Empty cells stay not listed.

Releases

Newest first

Open-weight releases, newest day first
GGUF 16 24 32 2×32
2026-10-15
10-15 tencent 30.06B / 3B active llama.cpp · vLLM · SGLang none 17 no fits fits fits
2026-10-14
10-14 black-forest-labs 9.41B llama.cpp · vLLM · SGLang none 5.3 fits fits fits fits
10-14 ibm-granite 29.99B llama.cpp none 16.9 no fits fits fits
10-14 ibm-granite 8.78B llama.cpp none 5 fits fits fits fits
10-14 microsoft 4.54B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 2.6 fits fits fits fits
2026-10-13
10-13 inclusionAI 572.12B / 8.94B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 323 no no no no
2026-10-12
10-12 LiquidAI 23.97B / 2B active llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 13.5 tight fits fits fits
2026-10-09
10-09 togethercomputer 14.77B llama.cpp · vLLM · SGLang none 8.3 fits fits fits fits
10-09 togethercomputer 8.19B llama.cpp · vLLM · SGLang none 4.6 fits fits fits fits
10-09 LiquidAI 23.97B / 2B active llama.cpp · vLLM · Ollama · SGLang none 13.5 tight fits fits fits
10-09 tencent 5.33B unknown none 3 fits fits fits fits
2026-10-07
10-07 reflection 501.27B / 23B active llama.cpp · Ollama · LM Studio · Lemonade yes 283 no no no no
10-07 openbmb 35.21B / 3B active llama.cpp · vLLM · SGLang none 19.9 no fits fits fits
2026-10-06
10-06 openbmb 1.55B llama.cpp · vLLM · SGLang none 0.9 fits fits fits fits
2026-10-05
10-05 LiquidAI 3.12B llama.cpp · vLLM · Ollama · SGLang none 1.8 fits fits fits fits
10-05 LiquidAI 0.59B Ollama none 0.3 fits fits fits fits
2026-10-02
10-02 Aleph-Alpha 78.1B / 1.22B active vLLM none 44.1 no no no fits
10-02 amazon 8.19B llama.cpp · vLLM · SGLang none 4.6 fits fits fits fits
10-02 amazon 2.03B llama.cpp · vLLM · SGLang none 1.1 fits fits fits fits
2026-09-29
09-29 microsoft 33.36B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 18.8 no fits fits fits
09-29 microsoft 8.77B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 5 fits fits fits fits
09-29 BAAI 27.36B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 15.4 no fits fits fits
09-29 internlm 35.96B llama.cpp · vLLM · SGLang none 20.3 no fits fits fits
2026-09-26
09-26 internlm 4.54B llama.cpp · vLLM · SGLang none 2.6 fits fits fits fits
09-26 internlm 0.85B llama.cpp · vLLM · SGLang none 0.5 fits fits fits fits
09-26 internlm 2.21B llama.cpp · vLLM · SGLang none 1.2 fits fits fits fits
2026-09-25
09-25 togethercomputer 0.87B llama.cpp · vLLM · Ollama · SGLang none 0.5 fits fits fits fits
2026-09-23
09-23 togethercomputer 4.66B llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 2.6 fits fits fits fits
2026-09-21
09-21 XiaomiMiMo 310.76B / 9.71B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 175.5 no no no no
09-21 XiaomiMiMo 1024.22B / 21.34B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 578.3 no no no no
09-21 apple 9.41B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 5.3 fits fits fits fits
09-21 XiaomiMiMo 9.41B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 5.3 fits fits fits fits
2026-09-20
09-20 JetBrains 12.15B / 2.5B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 6.9 fits fits fits fits
2026-09-16
09-16 EleutherAI 7.3B llama.cpp · vLLM none 4.1 fits fits fits fits
09-16 EleutherAI 7.3B llama.cpp · vLLM none 4.1 fits fits fits fits
2026-09-15
09-15 JetBrains 27.78B llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 15.7 no fits fits fits
2026-09-13
09-13 internlm 403.42B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 227.8 no no no no
2026-09-10
09-10 deepseek-ai 763.21B vLLM none 430.9 no no no no
09-10 allenai 8B llama.cpp · vLLM · SGLang none 4.5 fits fits fits fits
2026-09-08
09-08 CohereLabs 3.35B llama.cpp · vLLM · SGLang none 1.9 fits fits fits fits
2026-09-06
09-06 openbmb 2.52B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 1.4 fits fits fits fits
2026-09-03
09-03 BAAI 4.44B llama.cpp · vLLM · SGLang none 2.5 fits fits fits fits
2026-09-02
09-02 CohereLabs 3.35B llama.cpp · vLLM · SGLang none 1.9 fits fits fits fits
09-02 CohereLabs 3.35B llama.cpp · vLLM · SGLang none 1.9 fits fits fits fits
2026-08-31
08-31 deepseek-ai 304.65B / 7.14B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 172 no no no no
2026-08-27
08-27 Qwen 4.54B Ollama none 2.6 fits fits fits fits
08-27 tencent 14.77B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 8.3 fits fits fits fits
08-27 tencent 8.19B llama.cpp · vLLM · SGLang none 4.6 fits fits fits fits
08-27 tencent 7.94B llama.cpp · vLLM · SGLang none 4.5 fits fits fits fits
08-27 tencent 779.96B / 24.37B active llama.cpp · vLLM · SGLang none 440.4 no no no no
2026-08-26
08-26 allenai 1.47B unknown none 0.8 fits fits fits fits
08-26 allenai 7.63B unknown none 4.3 fits fits fits fits
08-26 allenai 8.31B unknown none 4.7 fits fits fits fits
08-26 allenai 8.31B unknown none 4.7 fits fits fits fits
08-26 allenai 8.25B unknown none 4.7 fits fits fits fits
08-26 allenai 8.25B unknown none 4.7 fits fits fits fits
2026-08-25
08-25 zai-org 321.32B llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 181.4 no no no no
08-25 zai-org 753.33B / 23.54B active llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 425.3 no no no no
2026-08-24
08-24 Qwen 180B llama.cpp · Strata · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 101.6 no no no no
2026-08-21
08-21 EleutherAI 0.12B unknown none 0.1 fits fits fits fits
2026-08-13
08-13 deepseek-ai 1650.5B / 25.79B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 931.9 no no no no
2026-08-10
08-10 CohereLabs 2.48B vLLM none 1.4 fits fits fits fits
2026-08-08
08-08 Qwen 2446.18B / 47.78B active llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 1381.1 no no no no
2026-08-06
08-06 skt 3.02B vLLM · SGLang none 1.7 fits fits fits fits
2026-08-05
08-05 Qwen 27.78B llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 15.7 no fits fits fits
2026-07-31
07-31 meituan-longcat 69.13B / 3.24B active unknown none 39 no no no fits
07-31 deepseek-ai 304.18B / 7.13B active llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 171.7 no no no no
2026-07-29
07-29 internlm 35.96B vLLM · SGLang none 20.3 no fits fits fits
07-29 LGAI-EXAONE 749.36B / 37B active llama.cpp · vLLM · SGLang none 423.1 no no no no
2026-07-28
07-28 skt 691.69B / 21.62B active unknown none 390.5 no no no no
07-28 LiquidAI 2.7B llama.cpp · vLLM · Ollama · LM Studio · Lemonade · SGLang yes 1.5 fits fits fits fits
2026-07-25
07-25 microsoft 4.74B unknown none 2.7 fits fits fits fits
2026-07-24
07-24 kakaocorp 1.29B unknown none 0.7 fits fits fits fits
07-24 kakaocorp 1.29B unknown none 0.7 fits fits fits fits
07-24 kakaocorp 3.18B llama.cpp · vLLM · SGLang none 1.8 fits fits fits fits
07-24 kakaocorp 3.51B llama.cpp · vLLM · SGLang none 2 fits fits fits fits
2026-07-23
07-23 BAAI 122.56B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 69.2 no no no no
07-23 BAAI 4.54B llama.cpp · vLLM · LM Studio · Lemonade · SGLang yes 2.6 fits fits fits fits
2026-07-22
07-22 upstage 250.29B / 6.26B active Ollama none 141.3 no no no no

This model

Hy-Memory-30B-A3B

tencent · 2026-10-15

Maker
tencent
Released
2026-10-15
Params
30.06B / 3B active
Architecture
HYV3ForCausalLM
Context
262,144
License
apache-2.0
GGUF
none listed
Q4_K_M
17 GiB estimate
16 / 24 / 32 / 2×32
no · fits · fits · fits

Runs on

Open on Hub

Sources

Hugging Face Hub public API. GGUF filenames from unsloth, bartowski, lmstudio-community, and ggml-org. Runs-on from llama.cpp convert architecture names or a listed GGUF pack, the vLLM supported-architectures registry, Ollama library slugs, Strata Qwen 3.8 Flash notes on /backends, and SGLang architecture names when listed. LM Studio and Lemonade follow GGUF. MLX is excluded. Not llama-bench.

Next steps

Real pages on this site. We do not invent a next click.

Last updated

esc

Type to filter. Enter opens the first match. Full-text search lives on /search.