batiai/ qwen3.5-35b

47 Downloads Updated 5 days ago

tools thinking

ollama run batiai/qwen3.5-35b:iq4

curl http://localhost:11434/api/chat \
  -d '{
    "model": "batiai/qwen3.5-35b:iq4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='batiai/qwen3.5-35b:iq4',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'batiai/qwen3.5-35b:iq4',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code

Claude Code ollama launch claude --model batiai/qwen3.5-35b:iq4

Codex

Codex ollama launch codex --model batiai/qwen3.5-35b:iq4

OpenCode

OpenCode ollama launch opencode --model batiai/qwen3.5-35b:iq4

OpenClaw

OpenClaw ollama launch openclaw --model batiai/qwen3.5-35b:iq4

Models

Name

1 model

Size

Context

Input

qwen3.5-35b:iq4

19GB · 256K context window · Text · 5 days ago

qwen3.5-35b:iq4

19GB

256K

Text

Readme

Qwen 3.5 35B-A3B — Quantized by BatiAI

Quantized from official Alibaba weights. Verified on real Mac hardware.

Models

Tag	Size	VRAM	M4 Max (128GB)	Use Case
iq4	17GB	23GB	26.6 t/s	36GB+ Mac

Quick Start

ollama run batiai/qwen3.5-35b:iq4

Why Qwen 3.5 35B-A3B?

MoE architecture — 35B total, only 3B active per token
Faster than 27B Dense despite being “larger” (26.6 vs 17.0 t/s)
Less VRAM than 27B (23GB vs 28GB) — MoE advantage
262K context window
Excellent Korean + tool calling + coding
Apache 2.0 license

Why MoE beats Dense

	35B-A3B (MoE)	27B (Dense)
Total params	35B	27B
Active params	3B	27B
VRAM	23GB	28GB
Speed	26.6 t/s	17.0 t/s

MoE only activates 3B params per token — 9x less compute than 27B Dense. Same quality, much faster.

RAM Requirements

Your Mac RAM	IQ4 (17GB)
16GB	❌
32GB	⚠️ Tight (23GB VRAM)
36GB+	✅ Fits
48GB+	✅ Fast
128GB	26.6 t/s

Full BatiAI Qwen 3.5 Lineup

Model	Size	VRAM	Speed (M4 Max)	Min Mac
batiai/qwen3.5-9b:q4	5.2GB	~8GB	12.5 t/s	16GB
batiai/qwen3.5-27b:iq4	14GB	28GB	17.0 t/s	32GB
batiai/qwen3.5-35b:iq4	17GB	23GB	26.6 t/s	36GB

For 36GB+ Mac, the 35B MoE is the clear winner — faster and less VRAM than 27B.

Why BatiAI?

Quantized directly from official Alibaba weights
IQ4_XS with imatrix — best quality at this size
Verified on MacBook Pro M4 Max (128GB)
Korean, tool calling, JSON generation all tested

Built for BatiFlow

Free, on-device AI automation for Mac. 5MB app, 100% local, unlimited.

https://flow.bati.ai

# Qwen 3.5 35B-A3B — Quantized by BatiAI

Quantized from official Alibaba weights. Verified on real Mac hardware.

## Models

| Tag | Size | VRAM | M4 Max (128GB) | Use Case |
|-----|------|------|----------------|----------|
| **iq4** | **17GB** | **23GB** | **26.6 t/s** | **36GB+ Mac** |

## Quick Start

```
ollama run batiai/qwen3.5-35b:iq4
```

## Why Qwen 3.5 35B-A3B?

- MoE architecture — 35B total, only 3B active per token
- **Faster than 27B Dense** despite being "larger" (26.6 vs 17.0 t/s)
- **Less VRAM than 27B** (23GB vs 28GB) — MoE advantage
- 262K context window
- Excellent Korean + tool calling + coding
- Apache 2.0 license

## Why MoE beats Dense

| | 35B-A3B (MoE) | 27B (Dense) |
|---|---|---|
| Total params | 35B | 27B |
| **Active params** | **3B** | **27B** |
| VRAM | **23GB** | 28GB |
| Speed | **26.6 t/s** | 17.0 t/s |

MoE only activates 3B params per token — 9x less compute than 27B Dense. Same quality, much faster.

## RAM Requirements

| Your Mac RAM | IQ4 (17GB) |
|-------------|-----------|
| 16GB | ❌ |
| 32GB | ⚠️ Tight (23GB VRAM) |
| 36GB+ | ✅ Fits |
| 48GB+ | ✅ Fast |
| 128GB | 26.6 t/s |

## Full BatiAI Qwen 3.5 Lineup

| Model | Size | VRAM | Speed (M4 Max) | Min Mac |
|-------|------|------|----------------|---------|
| batiai/qwen3.5-9b:q4 | 5.2GB | ~8GB | 12.5 t/s | **16GB** |
| batiai/qwen3.5-27b:iq4 | 14GB | 28GB | 17.0 t/s | **32GB** |
| **batiai/qwen3.5-35b:iq4** | **17GB** | **23GB** | **26.6 t/s** | **36GB** |

For 36GB+ Mac, the 35B MoE is the clear winner — faster and less VRAM than 27B.

## Why BatiAI?

- Quantized directly from official Alibaba weights
- IQ4_XS with imatrix — best quality at this size
- Verified on MacBook Pro M4 Max (128GB)
- Korean, tool calling, JSON generation all tested

## Built for BatiFlow

Free, on-device AI automation for Mac. 5MB app, 100% local, unlimited.

https://flow.bati.ai

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)