16 Apr 2026
Qwen3.6-35B-A3B
A smaller expert store for coding-agent experiments
35B reported total · text + image + video
Qwen / MoE
Qwen3-Coder-Next
Executable feedback matters as much as model size
New to model sizes or GPU memory? Start with weights, parameters and quantization →
Coder-Next is a strong candidate for evaluating local coding agents that can afford an 80B expert store. Its 3B active path makes it an instructive comparison with both small dense models and trillion-parameter coding systems.
80Breported total parameters
3Bactive parameters per token
262,144configured token positions
Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Evaluate it for
Code tools that can run tests, inspect repository state and measure completed changes.
Choose another path when
A device chosen using only the 3B active label or a workflow requiring native image input.
Release context
Why this release matters
Qwen built it on the hybrid Qwen3-Next base and emphasized executable coding tasks, environment interaction and reinforcement learning. The architecture and the agent-training process both explain its position.
The official blog displays February 2; its citation access date is February 3. These are different events.
Deployment starting points
Which hardware, how many, at what precision?
Start with an exact artifact and an engine that supports it. A configuration below is evidence of a documented path; its device count does not promise a particular throughput or concurrency.
No certified hardware configuration in this edition
A device chosen using only the 3B active label or a workflow requiring native image input.
The pinned model card establishes the architecture. It does not, by itself, prove that a particular GPU count runs this checkpoint at your target context. The capacity screen below helps rule out undersized allocations; a successful load and workload test are still required.
159.35 GB of tensor data in the pinned index
This is the index’s reported tensor payload, in decimal GB. It excludes file headers and runtime memory. Mixed precision, conversion and host/device placement determine how much GPU memory the loaded model needs.
Inspect the exact index metadata →Capacity screen: would the weights fit?
This estimates inference memory for a hypothetical uniform precision. It is useful for rejecting an allocation that is too small. It does not establish a working deployment or estimate training memory.
| Representation | Approximate weight floor | What remains to budget |
|---|---|---|
| 16-bit uniform | 160 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
| 8-bit uniform | 80 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
| 4-bit uniform | 40 GB | Cache, modality state, workspace, quantization metadata and uneven sharding |
Enter the total across the whole model replica, not per GPU: KV or recurrent cache for all concurrent requests, image/audio state and execution workspace. There is no universal default. We additionally leave 15% of each GPU unused.
160 GB for hypothetical weights
Enter a request-and-buffer budget to calculate the aggregate capacity requirement.
A count of one does not prove the engine supports this GPU. A count above one does not prove the model shards evenly or communicates efficiently. Publisher-reported, rounded model count; not an exact tensor census. Multimodal components and packaging can change the artifact size.
Architecture in practice
What the serving engine has to do
Two kinds of layers retain two kinds of state.
Hybrid attention does not mean zero request state.
A full-attention formula cannot describe every layer.
Mechanism schematic based on the pinned model card and configuration. It explains a design principle; it is not a full implementation graph.
Inspect the full checkpoint specifications
| Exact checkpoint | Qwen/Qwen3-Coder-Next |
|---|---|
| Text model type | qwen3_next |
| Layers | 48 |
| Hidden width | 2,048 |
| Attention / KV heads | 16 / 2 |
| Head dimension | 256 |
| Routed / selected experts | 512 / 10 |
| Configured positions | 262,144 |
| Documented extension | Not recorded for this checkpoint |
| Inputs → output | text → text |
| License metadata | apache-2.0 |
Open weights do not imply unrestricted use. Read the applicable license terms linked from the pinned card.
A useful comparison
What else belongs on the shortlist?
The 35B MoE offers a smaller multimodal store; K2.7-Code tests whether much larger residency earns more completed tasks.
12 Jun 2026
Kimi-K2.7-Code
A coding upgrade without a new size class
1,000B reported total · text + image + video
These are editorial comparison candidates. We have not run a matched quality or serving benchmark, so this is not a ranking.
Prepare to explain it
The interview lesson
Compare stored experts with activated experts, then explain why executable task feedback changes evaluation design.
Attention layouts · Quantization · Courses and worked examples
For a deployment evaluation, record exact weights, precision, engine, device count, interconnect, prompt/output lengths and concurrency. Report task success, errors, TTFT, TPOT and useful throughput together.
Sources and citation
Our assessment is editorial judgment based on the linked architecture and deployment evidence. Model facts are publisher-reported or attributed to runtime maintainers; no independent GPU benchmark was run.
AI Infra Interviews, “Qwen3-Coder-Next: release, architecture and deployment”, checked 2026-09-13. Preserve this date and the exact checkpoint when citing.