BLK · MODEL RELEASE TRACKER
What just dropped.
Newest local AI models, date-sorted. Every row carries quick fit verdicts for the four VRAM classes operators ask about — so you know in one glance whether to bother downloading a model before it starts loading. 60 models indexed.
For broader ecosystem news see /pulse. For the recommendation engine see /choose-my-gpu.
60 models shown · newest first
| Added | Model | Params | 8GB | 16GB | 24GB | 48GB | 96GB+ |
|---|---|---|---|---|---|---|---|
| 4d ago | Qwen 2.5 Coder 7B Instruct qwen · released 2024-11-12 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 3.5 235B-A17B (MoE) qwen · released 2026-05-01 | 397B | ✗ | ✗ | ✗ | ✗ | ✗ |
| 1w ago | DeepSeek V4 Pro (1.6T MoE) deepseek · released 2026-04-24 | 1600B | ✗ | ✗ | ✗ | ✗ | ✗ |
| 1w ago | Qwen 2.5-VL 3B qwen · released 2025-01-26 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | DeepSeek Coder V2 236B deepseek · released 2024-06-17 | 236B | ✗ | ✗ | ✗ | ✗ | △ |
| 1w ago | Qwen 2.5 3B Instruct qwen · released 2024-09-19 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2.5 1.5B Instruct qwen · released 2024-09-19 | 1.5B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2.5 0.5B Instruct qwen · released 2024-09-19 | 500M | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | DeepSeek V2.5 236B deepseek · released 2024-09-05 | 236B | ✗ | ✗ | ✗ | ✗ | △ |
| 1w ago | BGE Reranker v2 M3 other · released 2024-04-15 | 570M | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 3 Embedding 8B qwen · released 2025-06-05 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | NV-Embed v2 other · released 2024-09-09 | 7.85B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | BGE M3 other · released 2024-01-30 | 570M | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | DBRX Base dbrx · released 2024-03-27 | 132B | ✗ | ✗ | ✗ | ✗ | ✓ |
| 1w ago | EXAONE 3.5 2.4B exaone · released 2024-12-09 | 2.4B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | InternLM 2.5 7B Chat internlm · released 2024-07-03 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Falcon 3 7B Instruct falcon · released 2024-12-17 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | PaliGemma 2 10B gemma · released 2024-12-05 | 10B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | PaliGemma 2 3B gemma · released 2024-12-05 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | GLM-4V 9B glm · released 2024-06-04 | 13.9B | △ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Nemotron Mini 4B Instruct other · released 2024-09-13 | 4B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | SmolLM 2 1.7B Instruct other · released 2024-11-01 | 1.7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | SmolLM 2 360M Instruct other · released 2024-11-01 | 360M | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Tulu 3 70B other · released 2024-11-21 | 70B | ✗ | ✗ | ✗ | ~ | ✓ |
| 1w ago | Tulu 3 8B other · released 2024-11-21 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Granite 3.0 8B Instruct granite · released 2024-10-21 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Granite 3.0 2B Instruct granite · released 2024-10-21 | 2B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Jamba 1.5 Large other · released 2024-08-22 | 398B | ✗ | ✗ | ✗ | ✗ | ✗ |
| 1w ago | Jamba 1.5 Mini other · released 2024-08-22 | 52B | ✗ | ✗ | △ | ✓ | ✓ |
| 1w ago | Molmo 72B other · released 2024-09-25 | 72B | ✗ | ✗ | ✗ | ~ | ✓ |
| 1w ago | Molmo 7B-D other · released 2024-09-25 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | InternVL 2.5 78B other · released 2024-12-05 | 78B | ✗ | ✗ | ✗ | ~ | ✓ |
| 1w ago | InternVL 2.5 26B other · released 2024-12-05 | 26B | ✗ | △ | ✓ | ✓ | ✓ |
| 1w ago | LLaVA-OneVision 7B other · released 2024-08-06 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | LLaVA 1.6 Mistral 7B other · released 2024-01-30 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Whisper Large v3 Turbo other · released 2024-10-01 | 810M | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Whisper Large v3 other · released 2023-11-06 | 1.55B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | CodeQwen 1.5 7B qwen · released 2024-04-16 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | StarCoder 2 15B other · released 2024-02-28 | 15B | △ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | StarCoder 2 7B other · released 2024-02-28 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | StarCoder 2 3B other · released 2024-02-28 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Aya 23 35B other · released 2024-05-23 | 35B | ✗ | ✗ | ~ | ✓ | ✓ |
| 1w ago | Aya 23 8B other · released 2024-05-23 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Mistral Saba 24B mistral · released 2025-02-17 | 24B | ✗ | ~ | ✓ | ✓ | ✓ |
| 1w ago | Ministral 8B Instruct mistral · released 2024-10-16 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Ministral 3B Instruct mistral · released 2024-10-16 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2-VL 7B qwen · released 2024-08-29 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2.5 Math 72B qwen · released 2024-09-19 | 72B | ✗ | ✗ | ✗ | ~ | ✓ |
| 1w ago | Qwen 2.5 Math 7B qwen · released 2024-09-19 | 7B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2.5 Coder 3B qwen · released 2024-11-12 | 3B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Qwen 2.5 Coder 1.5B qwen · released 2024-11-12 | 1.5B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Llama 3.3 8B Instruct llama · released 2025-04-12 | 8B | ~ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Devstral Small 2 24B mistral · released 2025-09-25 | 24B | ✗ | ~ | ✓ | ✓ | ✓ |
| 1w ago | Phind CodeLlama 34B v2 llama · released 2023-09-01 | 34B | ✗ | △ | ~ | ✓ | ✓ |
| 1w ago | DeepSeek MoE 16B Base deepseek · released 2024-01-15 | 16B | ✗ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Baichuan 4 13B baichuan · released 2024-10-30 | 13B | △ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Phi-4 Reasoning Mini 4B phi · released 2026-04-08 | 3.8B | ✓ | ✓ | ✓ | ✓ | ✓ |
| 1w ago | Aya Expanse 32B command-r · released 2024-10-22 | 32B | ✗ | △ | ~ | ✓ | ✓ |
| 1w ago | DeepSeek R1 Distill Mistral 24B deepseek · released 2025-03-18 | 24B | ✗ | ~ | ✓ | ✓ | ✓ |
| 1w ago | Command R+ (Aug 2024) command-r · released 2024-08-30 | 104B | ✗ | ✗ | ✗ | △ | ✓ |
✓
Comfortable
Q4 fits with KV headroom
~
Tight
Q4 fits, small context
△
Marginal
IQ3 only, expect degradation
✗
Doesn't fit
Won't run usefully
HOW FIT VERDICTS ARE DERIVED
Quick footprint estimate at Q4_K_M: params × 0.6 GB + 1.5 GB runtime overhead. Comfortable means the rig has ≥1.4× headroom for KV cache and multi-turn context. Tight means it fits but you'll bump the ceiling on long context. Marginal means only aggressive (IQ3 / IQ2) quants work, with quality degradation. Doesn't fit means weights alone won't load without RAM offload, which crushes tok/s. Frontier-class models (400B+) render as no-fit on every single-rig VRAM class — they need multi-GPU or cloud.