Ling 3.0 Flash vs Ling 3.1 Flash

Wins 0 of 6 benchmarks
$1.00 / task
Wins 5 of 6 benchmarks
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Ling 3.0 Flash3.2
Ling 3.1 Flash61.7
GDPval-AA (work tasks) 2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work
Ling 3.0 Flash22.4
Ling 3.1 Flash55.9
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
Ling 3.0 Flash0
Ling 3.1 Flash33.3
AA-Briefcase 1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating
Ling 3.0 Flash797.6
Ling 3.1 Flash1,400
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
Ling 3.0 Flash73
Ling 3.1 Flash83
τ³-Banking · AA τ³-Banking · AA: Agents retrieve banking knowledge and execute tools to resolve customer-support workflows
Ling 3.0 Flash27.2
Ling 3.1 Flash—

Specs

SpecLing 3.0 FlashLing 3.1 Flash
Input, per 1M$0.06$0.30
Output, per 1M$0.18$0.90
Cost per task?—$1.00
Speed, tok/s182145
Context262K1M
WeightsOpenOpen
ReleasedAug 2026Oct 2026
Memory (4-bit)76 GB343 GB
LicenceMITMIT
Commercial useAllowedAllowed
See images——
Read PDFs——
Hear audio——
Watch video——
Call tools✓✓
Force a tool callPartialPartial
Return strict JSONPartial✓
Search the web??
Run code??
Use a computer??

What breaks when you switch

Ling 3.0 Flash → Ling 3.1 Flash
Memory (4-bit): 76 GB → 343 GB
Ling 3.1 Flash → Ling 3.0 Flash
Context: 1M → 262K