Ling 3.1 Flash
- Input
- $0.30/M tok
- Output
- $0.90/M tok
- Blended
- $0.45/M tok · 3:1
- Speed
- 213tok/s
- Context
- 262Ktokens
Rankings
Self-hostingEstimated
How we estimate →What it would cost to run these open weights yourself on rented GPUs, compared with the API price above.
- Hardware
- 8× B200 180GB
- FP8 weights
- Throughput
- ~31,380 tok/s
- many requests batched
- Price
- $0.18–1.61 /M tok
- busy → light use
- On your own machine
- Mac Studio M5 Ultra 512GB (int4, ~67 tok/s single-stream)
- single consumer GPU or Mac
Assumptions (5)
- FP8 weights (560B params, 25B active per token) + 25% KV-cache headroom ≈ 700 GB
- 8× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
- ~31,380 output tok/s aggregate at batch 898 (bandwidth-bound); MoE compute scales with active params
- Blended 3:1 input:output; prefill ~340,200 tok/s
- Low = 75% utilization at the low GPU price; high = 20% at the high price
Benchmark scores
~ italic, dashed = no published score yet, estimated from related benchmarks.
Coding
SWE-bench VerifiedEstimated
~#2 of 20 (estimated)~85.5%
No published score — estimated from related benchmarks
SWE-bench Pro (Public)Estimated
~#6 of 23 (estimated)~67.3%
- #1Claude Opus 5.589.9%
- #2Claude Sonnet 5.581.3%
- #5Intern-S2-397B68.5%
- ~#6Ling 3.1 FlashEstimated67.3%
- #6Beam 1.0 501B A23B66.1%
No published score — estimated from related benchmarks
DeepSWE v1.1Estimated
~#28 of 43 (estimated)~61.6%
- #1Gemini 4 Argon78.8%
- #2DeepSeek V4.1 Flash 101777.0%
- #27Mistral Large 461.7%
- ~#28Ling 3.1 FlashEstimated61.6%
- #28GLM-5.361.4%
No published score — estimated from related benchmarks
Terminal-Bench 4.0Verified
#17 of 9833.3%
- #1Kimi K3.1 Pro67.2%
- #2Claude Sonnet 5.563.6%
- #16DeepSeek V4.1 Flash 101733.8%
- #17Ling 3.1 Flash33.3%
- #17Muse Spark 1.333.3%
artificialanalysis.ai · 2026-10-15
LiveCodeBenchEstimated
~#10 of 28 (estimated)~85.0%
- #1Qwen3.8-Omni-Flash92.6%
- #2Qwen3.8 Flash-Next91.9%
- #9Nemotron 3.5 Super VL NVFP487.4%
- ~#10Ling 3.1 FlashEstimated85.0%
- #10Ternary Bonsai 27B82.8%
No published score — estimated from related benchmarks
SciCodeVerified
#31 of 11754.1%
- #1Claude Opus 5.566.9%
- #2Claude Fable 5.163.1%
- #30Mistral Large 454.2%
- #31Ling 3.1 Flash54.1%
- #31Qwen3.8 2.4T A95B54.1%
artificialanalysis.ai · 2026-10-15
AA Coding Agent IndexEstimated
~#17 of 24 (estimated)~45.5
- #1Claude Opus 5.561.7
- #2Claude Opus 559.7
- #16Muse Spark 1.345.7
- ~#17Ling 3.1 FlashEstimated45.5
- #17DeepSeek V4 Pro (0813)43.1
No published score — estimated from related benchmarks
Hard reasoning
GPQA DiamondEstimated
~#5 of 191 (estimated)~93.6%
- #1GPT-6 Astra96.1%
- #2Gemini 3.8 Flash95.3%
- #4Claude Fable 5.193.7%
- ~#5Ling 3.1 FlashEstimated93.6%
- #5GPT-5.593.5%
No published score — estimated from related benchmarks
Humanity's Last ExamVerified
#34 of 19739.4%
- #1Claude Opus 5.561.4%
- #2Claude Fable 5.159.1%
- #33GLM-5.3 Flash39.9%
- #34Ling 3.1 Flash39.4%
- #35DeepSeek V4.1 Flash39.2%
artificialanalysis.ai · 2026-10-15
AIME (latest)Estimated
~#2 of 33 (estimated)~98.2%
No published score — estimated from related benchmarks