Flint vs Ling 3.1 Flash

Overall
21
vs
Overall
45

Ling 3.1 Flash is clearly better. It wins 47 of the 51 test cases where they differ.

Learn to build evals like this in the AI Engineering Academy

Task by task

A score turns orange only when that model wins clearly more of the task's test cases. Under each score, the average time per test case on the machine it ran on. Flint spends 1.7× as many thinking tokens per test case.

Thinking tokensAnswer tokens
TaskAB
Overall
211m 17s
10.9K, 90% thinking7.9K, 73% thinking
451m 16s
Agents
1447.7s
4.8K, 85% thinking4.5K, 66% thinking
4155.9s
Structured Extraction
291m 17s
10.9K, 87% thinking8.1K, 69% thinking
711m 11s
Coding
141m 19s
12K, 98% thinking11.3K, 89% thinking
241m 42s
Grounded Answers
2957.7s
8.7K, 93% thinking6.7K, 74% thinking
5258.9s
Instruction Following
352m 28s
22.2K, 93% thinking11.4K, 77% thinking
761m 55s
Business Logic
101m 20s
11.7K, 78% thinking9.7K, 56% thinking
381m 22s
Security & permissions
3552.6s
7.1K, 90% thinking5.2K, 74% thinking
5456.1s
Diagnosis & investigation
057.0s
7.3K, 99% thinking3.2K, 72% thinking
042.1s

What you'd run

The build behind each score. Orange marks the faster, smaller, cheaper or longer-context side.

SpecAB
Speed
149 tok/s (better)
112 tok/s
Machine
Hosted
Hosted
Memory at 32K
—
—
Context
256K
256K
Cost per run
Not measured
Not measured
Cost per case
Not measured
Not measured
Time per case
1m 17s
1m 16s (better)
Quantization
BF16
—
Harness
openrouter
openrouter

The race

No lead changes

Each model's running score as the test cases go by. The dashed line is a perfect run.

FlintLing 3.1 Flash
0255075100501001502145

Every test case

153 test cases

Each square is one test case, colored by the model that scored higher.

4 won by Flint47 won by Ling 3.1 Flash102 tied