Inference
EN
InclusionAI

Ling 3.0 Flash

Ling 3.0 Flash is a reasoning model from InclusionAI in the Ling 3.0 family. 25 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-07
RankedEnough results for an index.
IndexAll categories, out of 100.50.9±4.7Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.73/sMedian 67/s
First tokenSeconds to the first output token.0.9 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$0.021Median $0.60
Output / 1MUS dollars per 1M output tokens.$0.063Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.—Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
inclusionai/ling-3.0-flash
Max outputMaximum output tokens per request.
33K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
No
Cache read / 1MUS dollars per 1M cached tokens.
<$0.01
WeightsLink to the published weights.
Hugging Face (opens in a new tab) ↗
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
2 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
OpenRouterRouter$0.021$0.063262K
Novita AIHost$0.06$0.18262K

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Run locally

127.49B parameters
16-bit
268.7 GB
8-bit
143.2 GB
4-bit
82.2 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.8 counted53.5109 / 218
CodingCode writing and repair.7 counted50.0137 / 209
ReasoningLogic problems and puzzles.2 counted48.8124 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.4 counted46.1163 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted58.6—
MathMath problems.3 counted54.567 / 168

Results

25 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Berkeley Function Calling Leaderboard v4Agentic73.0%5 / 26
LiveCodeBenchCoding84.0% ±1.041 / 142
GPQA Diamond · Vals AIKnowledge84.8% ±2.249 / 135
Graduate-Level Google-Proof Q&AKnowledge85.0%53 / 141
Artificial Analysis GPQA DiamondKnowledge85.5%75 / 199
Artificial Analysis Humanity's Last ExamKnowledge23.7%102 / 222
Artificial Analysis Long Context ReasoningReasoning73.0%103 / 217
Critical Physics TasksReasoning1.7%109 / 214
Scientific Code BenchmarkCoding41.2%18 / 35
Harvard-MIT Mathematics Tournament February 2026Math87.0%14 / 27
Instruction Following BenchmarkInstruction74.5%31 / 57
SWE-bench ProCoding56.6%47 / 83
MMLU ProKnowledge82.0% ±0.480 / 135
GDPval-AA normalizedAgentic22.4%79 / 133
IMOAnswerBenchMath83.7%6 / 10
AIME 2026Math93.2%22 / 34
BrowseCompAgentic72.2%41 / 63
MCP AtlasAgentic65.5%34 / 52
WideResearchAgentic73.6%11 / 16
Artificial Analysis SciCodeCoding42.0%80 / 114
GPQA Diamond · BenchLMKnowledge85.0%59 / 84
Terminal-Bench 2.1 · BenchLMAgentic57.0%66 / 86
Terminal-Bench 2.1 · Vals AIAgentic50.2% ±2.558 / 75
Humanity's Last ExamKnowledge22.7%61 / 77
SWE-benchCoding65.2% ±2.172 / 87
Artificial Analysis Omniscience AccuracyKnowledge18.2%140 / 164
SkillsBenchCoding27.9% ±3.833 / 35
Vibe Code Bench v1.1Coding2.9% ±2.1116 / 123
Artificial Analysis Tau3-BankingAgentic28.0%22 / 22
Hermes Index: agentic capability and cost per task inside Hermes AgentAgentic21.6%16 / 16
Code MigrationCoding0.0%77 / 77
CorpFin v2Finance61.9% ±1.049 / 131
MedScribeHealthcare80.9% ±2.054 / 109
TaxEval v2Finance70.6% ±0.989 / 143
LegalBenchLegal79.7% ±0.593 / 149
MedCodeHealthcare32.3% ±1.989 / 106
Finance Agent (v2)Finance30.3% ±0.371 / 79
EMBFinance25.0% ±2.568 / 75
Public Benefits Bench v1.1Public benefits50.7% ±1.346 / 50
Legal Research BenchLegal0.0%90 / 90
LiveCodeBench v5Coding82.8%—
Data Research and Analysis with Complex OperationsAgentic70.4%—
Hermes Bench: 150 agent tasks graded on the files and state the agent leaves behindAgentic53.8%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedmodels.devMIT

History

5 releases
Index51.7 → 50.9
23 Oct 202612 Nov 2026
Output / 1M$0.063 → $0.063
23 Oct 202612 Nov 2026

Questions

How much does Ling 3.0 Flash cost?
$0.021 per 1M input tokens, $0.063 per 1M output tokens.
How good is Ling 3.0 Flash?
Its index is 50.9 out of 100. That is place 143 of 283 ranked models.
What is Ling 3.0 Flash best at?
Math: 54.5 out of 100, place 67 of 168.
How fast is Ling 3.0 Flash?
73 output tokens per second, 0.9 s to the first token.
What is the context window of Ling 3.0 Flash?
262K tokens.
Is Ling 3.0 Flash open source?
Yes. Its weights are published.
When was Ling 3.0 Flash released?
July 2026.

Badge

Ling 3.0 Flash on Inference360: place 143 of 283, index 50.9
HTML
<a href="https://inference360.ai/models/ling-3-0-flash"><img src="https://inference360.ai/badges/ling-3-0-flash.svg" alt="Ling 3.0 Flash on Inference360: place 143 of 283, index 50.9" height="20"></a>
Markdown
[![Ling 3.0 Flash on Inference360: place 143 of 283, index 50.9](https://inference360.ai/badges/ling-3-0-flash.svg)](https://inference360.ai/models/ling-3-0-flash)

Compare with

GPT-5.4 nano50.9Nemotron 3.5 Super VL 120B A12B NVFP450.8Nemotron 3 Ultra50.8Nemotron 3.5 Super VL 120B A12B51.0dots3-note Preview63.4

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeNemotron 3 Ultra50.8 · Free

More from InclusionAI

All models
LLaDA2.2-mini24.3LLaDA2.2-flashUnrankedLing 3.0 Flash FP8UnrankedLing 3.0 TinyUnrankedLing 3.0 Flash FinUnranked