[object Object]

← back to Exo

Add Qwen3-Coder-Next model cards (#1367)

ffe6396c9145f90b6078c13367969010daef6047 · 2026-02-05 05:37:18 -0800 · Alex Cheema

## Motivation

Qwen3-Coder-Next just dropped on mlx-community in several quantizations.
It's an 80B MoE model (Qwen3NextForCausalLM) which we already have
tensor parallelism support for via QwenShardingStrategy — just needs
model cards.

## Changes

Added model cards for all 5 available quantizations:
- `mlx-community/Qwen3-Coder-Next-4bit` (~46GB)
- `mlx-community/Qwen3-Coder-Next-5bit` (~58GB)
- `mlx-community/Qwen3-Coder-Next-6bit` (~69GB)
- `mlx-community/Qwen3-Coder-Next-8bit` (~89GB)
- `mlx-community/Qwen3-Coder-Next-bf16` (~158GB)

All with `supports_tensor = true` since the architecture is already
supported.

## Why It Works

`Qwen3NextForCausalLM` is already handled by QwenShardingStrategy in
auto_parallel.py and is in the supports_tensor allowlist in
model_cards.py. No code changes needed — just the TOML card files.

## Test Plan

### Manual Testing
<!-- n/a - model card addition only -->

### Automated Testing
- `basedpyright` — 0 errors
- `ruff check` — passes
- `nix fmt` — no changes
- `pytest` — 173 passed, 1 skipped


🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

Files touched

Diff

commit ffe6396c9145f90b6078c13367969010daef6047
Author: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
Date:   Thu Feb 5 05:37:18 2026 -0800

    Add Qwen3-Coder-Next model cards (#1367)
    
    ## Motivation
    
    Qwen3-Coder-Next just dropped on mlx-community in several quantizations.
    It's an 80B MoE model (Qwen3NextForCausalLM) which we already have
    tensor parallelism support for via QwenShardingStrategy — just needs
    model cards.
    
    ## Changes
    
    Added model cards for all 5 available quantizations:
    - `mlx-community/Qwen3-Coder-Next-4bit` (~46GB)
    - `mlx-community/Qwen3-Coder-Next-5bit` (~58GB)
    - `mlx-community/Qwen3-Coder-Next-6bit` (~69GB)
    - `mlx-community/Qwen3-Coder-Next-8bit` (~89GB)
    - `mlx-community/Qwen3-Coder-Next-bf16` (~158GB)
    
    All with `supports_tensor = true` since the architecture is already
    supported.
    
    ## Why It Works
    
    `Qwen3NextForCausalLM` is already handled by QwenShardingStrategy in
    auto_parallel.py and is in the supports_tensor allowlist in
    model_cards.py. No code changes needed — just the TOML card files.
    
    ## Test Plan
    
    ### Manual Testing
    <!-- n/a - model card addition only -->
    
    ### Automated Testing
    - `basedpyright` — 0 errors
    - `ruff check` — passes
    - `nix fmt` — no changes
    - `pytest` — 173 passed, 1 skipped
    
    
    🤖 Generated with [Claude Code](https://claude.com/claude-code)
    
    Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
---
 .../mlx-community--Qwen3-Coder-Next-4bit.toml                     | 8 ++++++++
 .../mlx-community--Qwen3-Coder-Next-5bit.toml                     | 8 ++++++++
 .../mlx-community--Qwen3-Coder-Next-6bit.toml                     | 8 ++++++++
 .../mlx-community--Qwen3-Coder-Next-8bit.toml                     | 8 ++++++++
 .../mlx-community--Qwen3-Coder-Next-bf16.toml                     | 8 ++++++++
 5 files changed, 40 insertions(+)

diff --git a/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-4bit.toml b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-4bit.toml
new file mode 100644
index 00000000..9e6d3511
--- /dev/null
+++ b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-4bit.toml
@@ -0,0 +1,8 @@
+model_id = "mlx-community/Qwen3-Coder-Next-4bit"
+n_layers = 48
+hidden_size = 2048
+supports_tensor = true
+tasks = ["TextGeneration"]
+
+[storage_size]
+in_bytes = 45644286500
diff --git a/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-5bit.toml b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-5bit.toml
new file mode 100644
index 00000000..ebf487af
--- /dev/null
+++ b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-5bit.toml
@@ -0,0 +1,8 @@
+model_id = "mlx-community/Qwen3-Coder-Next-5bit"
+n_layers = 48
+hidden_size = 2048
+supports_tensor = true
+tasks = ["TextGeneration"]
+
+[storage_size]
+in_bytes = 57657697020
diff --git a/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-6bit.toml b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-6bit.toml
new file mode 100644
index 00000000..ef801631
--- /dev/null
+++ b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-6bit.toml
@@ -0,0 +1,8 @@
+model_id = "mlx-community/Qwen3-Coder-Next-6bit"
+n_layers = 48
+hidden_size = 2048
+supports_tensor = true
+tasks = ["TextGeneration"]
+
+[storage_size]
+in_bytes = 68899327465
diff --git a/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-8bit.toml b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-8bit.toml
new file mode 100644
index 00000000..05b92116
--- /dev/null
+++ b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-8bit.toml
@@ -0,0 +1,8 @@
+model_id = "mlx-community/Qwen3-Coder-Next-8bit"
+n_layers = 48
+hidden_size = 2048
+supports_tensor = true
+tasks = ["TextGeneration"]
+
+[storage_size]
+in_bytes = 89357758772
diff --git a/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-bf16.toml b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-bf16.toml
new file mode 100644
index 00000000..93798231
--- /dev/null
+++ b/resources/inference_model_cards/mlx-community--Qwen3-Coder-Next-bf16.toml
@@ -0,0 +1,8 @@
+model_id = "mlx-community/Qwen3-Coder-Next-bf16"
+n_layers = 48
+hidden_size = 2048
+supports_tensor = true
+tasks = ["TextGeneration"]
+
+[storage_size]
+in_bytes = 157548627945

← 3a9baeb9 EXO: add CLI flags for root install/uninstall  ·  back to Exo  ·  feat: show download availability in model picker (#1377) 5c2f29f3 →