Weekly archive

New local AI models, Jul 20–26, 2026

7 open models released Jul 20–26, 2026, with how each runs on five reference devices.

  1. KAT-Coder V2.5 Dev (35B-A3B)

    Kuaishou · LLM · 35B (3.0B active) ·

    91Excellent quality (A+) · 91 of 100: how capable the model is at F16, whatever it runs on. Match score 87/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  2. Solar Open 2 250B

    Upstage · LLM · 250B (15.0B active) ·

    87Excellent quality (A+) · 87 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 91/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  3. LTX-2.5 (22B, video + audio)

    Lightricks · Video · 21.0B ·

    88Excellent quality (A+) · 88 of 100: how capable the model is at FP16, whatever it runs on. Match score 79/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  4. Nanbeige4.2 3B

    BOSS Zhipin (Nanbeige) · LLM · 4.2B ·

    68Good quality (B) · 68 of 100: how capable the model is at F16, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  5. Apertus v1.5 70B

    Swiss AI · LLM · 72B ·

    62Good quality (B) · 62 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 66/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  6. Mage-VL 4.7B

    Microsoft · LLM · 4.7B ·

    56Good quality (B) · 56 of 100: how capable the model is at F16, whatever it runs on. Match score 67/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

  7. Apertus v1.5 8B

    Swiss AI · LLM · 8.9B ·

    46Fair quality (C) · 46 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 63/100 (quality, fit, speed and memory use) sets the ranking.

    Runs on

    Memory, every quant and how to run it Check your own device

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.

Was this page helpful?