Qwen2.5 Coder 32B Instruct vs Qwen3.6 35B A3B

Wins 0 of 5 benchmarks
$0.48 / task
Wins 1 of 5 benchmarks
Different efforts: non-thinking · thinking
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Qwen2.5 Coder 32B Instruct3.5
Qwen3.6 35B A3B22.2
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Qwen2.5 Coder 32B Instruct—
Qwen3.6 35B A3B0
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Qwen2.5 Coder 32B Instruct—
Qwen3.6 35B A3B5.2
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Qwen2.5 Coder 32B Instruct1,207
Qwen3.6 35B A3B—
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Qwen2.5 Coder 32B Instruct—
Qwen3.6 35B A3B75

Specs

SpecQwen2.5 Coder 32B InstructQwen3.6 35B A3B
Input, per 1M$0.06$0.25
Output, per 1M$0.15$1.49
Cost per task?—$0.48
Speed, tok/s—138
Context131K262K
WeightsOpenOpen
ReleasedNov 2024Apr 2026
Memory (4-bit)20 GB22 GB
LicenceApache-2.0Apache-2.0
Commercial useAllowedAllowed
See images—✓
Read PDFs——
Hear audio——
Watch video—✓
Call tools—✓
Force a tool call—Partial
Return strict JSONPartial✓
Search the web??
Run code??
Use a computer??

What breaks when you switch

Qwen2.5 Coder 32B Instruct → Qwen3.6 35B A3B
Memory (4-bit): 20 GB → 22 GB
Qwen3.6 35B A3B → Qwen2.5 Coder 32B Instruct
Context: 262K → 131K
No image input