Skip to content
Model Pareto

Ling-3.0-flash-Fin

inclusionAI (Ant Group)Open weightshuggingface.co ↗
Input
$0.075/M tok
Output
$0.22/M tok
Blended
$0.111/M tok · 3:1
Speed
290tok/s
Context
262Ktokens
Cost per task
~$0.09/taskest.

Open weights (MIT); finance-tuned Ling-3.0-flash (124B total / 5.1B active MoE). Release date = OpenRouter listing; AA lists 2026-09-11. Price and speed are AA's median across hosts. Not ranked in Agents: its only agentic score is the finance-specific tau3-banking.

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
2× H200 141GB
FP8 weights
Throughput
~9,742 tok/s
many requests batched
Price
$0.079–0.65 /M tok
busy → light use
On your own machine
Mac Studio M5 Max 128GB (int4, ~170 tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • FP8 weights (124B params, 5.1B active per token) + 25% KV-cache headroom ≈ 155 GB
  • 2× H200 141GB at $3.59–$7.91/GPU-hour on-demand (2026-09-24)
  • ~9,742 output tok/s aggregate at batch 350 (bandwidth-bound); MoE compute scales with active params
  • Blended 3:1 input:output; prefill ~183,349 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

~#23 of 24 (estimated)~50.8%
  1. #1Claude Fable 5.293.0%
  2. #2Claude Opus 5.589.9%
  3. #22Muse Glimmer51.2%
  4. ~#23Ling-3.0-flash-FinEstimated50.8%
  5. #23Nex-N2.5-Mini43.8%
No published score — estimated from related benchmarks
~#25 of 26 (estimated)~30.0
  1. #1Claude Opus 5.561.7
  2. #2Claude Opus 559.7
  3. #24GPT-6 Luna31.1
  4. ~#25Ling-3.0-flash-FinEstimated30.0
  5. #25Muse Spark Open26.3
No published score — estimated from related benchmarks

Hard reasoning