HomeAI models
inclusionAI: Ling 3.0 Flash Sante (free)
•
Ling 3.0 Flash
•
Released 
September 2026


Ling 3.0 Flash

Health and medicine fine-tune of Ling 3.0 Flash with medical reasoning, evidence-based retrieval and function calling, free via API.

Modality:
Text
PDF
model ID
inclusionai/ling-3.0-flash-sante:free
Start for free
arrow black icon
View API Docs
arrow white icon

Output Speed *

382.49
tok/s

Intelligence Index *

24.9
/ 100

Context Window *

262144
tokens

Input price

0
Anytoken

Output price

0
Anytoken
A Medical-Domain Fine-Tune Built for Clinical Reasoning and Evidence Retrieval Ling 3.0 Flash Sante is inclusionAI's (Ant Group) health and medicine fine-tune of Ling 3.0 Flash. It keeps the base model's hybrid Mixture-of-Experts architecture — 124B total parameters with roughly 5.1B active per token, a 262,144-token context window, and function calling — while continued training targets medical knowledge reasoning, clinical safety, and evidence-based retrieval. It sits in the specialist tier of the Ling 3.0 family alongside the Fin (finance) variant. Teams building medical literature search, clinical decision-support prototypes and long-horizon medical question answering benefit most, provided outputs are reviewed by qualified professionals. Integrate Ling 3.0 Flash Sante via the AnyAPI.ai API

Performance

Where Ling 3.0 Flash Sante Concentrates Its Gains

Sante's differentiation is concentrated in medical knowledge and reasoning rather than general capability. On inclusionAI's own evaluations it reports leading results among flash-size models, scoring 53.88 on MedXpertQA-Text and 83.83 on DiagnosisArena-MCQ, plus safety-oriented scores such as 82.06 on MedEthicAlign. Because these are vendor-reported and no independent third-party medical benchmark exists yet, they indicate direction more than confirmed ranking. In production this means Sante is best treated as a research and workflow assistant for medical literature search, retrieval and summarization — with human clinical review — not as an autonomous diagnostic system.

Benchmarks

The Independent Benchmark Gap for Sante

As of release, no independent third-party benchmark exists specifically for Ling 3.0 Flash Sante — there is no Artificial Analysis page or public arena result for this exact fine-tune, and weights were not published on Hugging Face. Available medical scores (MedXpertQA-Text 53.88, DiagnosisArena-MCQ 83.83, AFUMED-Drug 89.59) are self-reported by inclusionAI. For context, the underlying Ling 3.0 Flash measured roughly 319.8 tokens/second output and a 2.54s time to first token on the provider's own API via Artificial Analysis, but those figures describe the base model, not this fine-tune. Treat all Sante-specific numbers as vendor claims pending independent verification.

Output Speed

*
382.49
tok/s

Intelligence Index

*
24.9
/ 100

MMLU *

Broad world knowledge and problem-solving
0
%

GPQA *

PhD-level scientific reasoning across physics, biology, chemistry.
86
%

HLE *

Adherence to multi-step structured instructions.
24
%

LiveCodeBench *

Tool-calling reliability in long agentic loops.
0
%

Technical Specifications

What the model supports

Ling 3.0 Flash Sante is a text-in, text-out Mixture-of-Experts model with 124B total parameters, roughly 5.1B active per token, and a 262,144-token context window supporting up to 32,768 output tokens. Thinking mode is enabled by default, and it accepts tools and tool_choice for function calling. The large context is the most production-relevant feature: it allows multi-document medical reasoning across guidelines, literature and patient records in a single request. Note it does not enforce response_format, so structured JSON output is not schema-guaranteed and must be validated downstream.
Verified Specifications — 
Ling 3.0 Flash
*
Input modalities
Text
PDF
output modalities
Text
Context window
262144
 tokens
Maximum output tokens
32768
Reasoning
Yes
Knowledge cutoff
September 2026
Pricing (standard)
0
 AnyTokens in
 / 
0
 AnyTokens out

Limitations & Trade-offs

Where Ling 3.0 Flash falls short

1
No independent verification. Ling 3.0 Flash Sante has no third-party benchmark, no Artificial Analysis page and no arena result; all medical scores are self-reported by inclusionAI. This matters most in healthcare, where the cost of a wrong answer is high and regulated settings demand auditability. If you need reproducible evidence of performance, the open-weight base Ling 3.0 Flash is a safer starting point.
2
Weights not released. Unlike the base model, Sante is a proprietary API checkpoint with no Hugging Face weights download, so it cannot be self-hosted, fine-tuned further, or independently audited. Teams with data-residency, offline, or on-premise requirements — common in clinical environments — cannot deploy this variant locally and must rely on hosted API access.
3
No enforced structured output. The model accepts function-calling tools but does not support response_format, so JSON output is not schema-guaranteed. Applications that depend on strictly validated structured responses — for example populating clinical data fields — must add their own parsing and validation layer rather than trusting native schema enforcement.
4
Not a medical device. inclusionAI and hosting platforms explicitly frame Sante as a developer API for research, retrieval and workflow assistance, not a diagnostic tool. Outputs should be reviewed by qualified professionals and never used directly as diagnosis, treatment instructions, or independent clinical decisions. This restricts it to assistive and prototyping roles rather than autonomous clinical deployment.

Best-Fit Workloads

Where this model earns its place

01

Medical literature search and synthesis

‍
The 262,144-token context lets Sante ingest multiple abstracts, guidelines, or review sections and produce grounded, evidence-based synthesis in a single request. Combined with function calling for retrieval tools, this fits research assistants and literature-review workflows. Because scores are vendor-reported and outputs require human review, treat results as drafts for expert verification rather than final answers.

02

Clinical question answering prototypes

‍
The fine-tune targets medical knowledge reasoning and clinical safety, with vendor-reported leadership among flash-size models on MedXpertQA-Text and DiagnosisArena-MCQ. This suits building and testing clinical decision-support prototypes and medical Q&A assistants. Keep human clinical oversight in the loop, since the model is positioned as assistive and lacks independent benchmark confirmation.

03

Evidence-based retrieval agents

‍
inclusionAI emphasizes deep research and evidence-based retrieval as a core focus, and function-calling support lets Sante orchestrate external search and grounding tools. This fits source-grounded medical agents that must cite reliable evidence. Note that without enforced structured outputs, citation and source formatting need downstream validation.

04

Long-horizon medical workflows

‍
With hybrid reasoning and a large context window, Sante is built for multi-step medical tasks that span retrieval, review and drafting rather than isolated single-turn answers. This supports workflow assistants that carry substantial patient or reference context across steps. Thinking mode adds latency and token use, so budget for slower responses on complex chains.

Pricing in anytokens via AnyAPI
Input
0
₳
Output
0
₳
Cache write
0
₳
Cache read
0
₳
Get API access
arrow black icon
Read Docs

on this page

PerformanceBenchmarksTechnical SpecificationsLimitations & Trade-offsBest-Fit WorkloadsIntegrationFAQ

Integration

Access Ling 3.0 Flash via AnyAPI.ai

Access Ling 3.0 Flash through AnyAPI.ai using a unified API built for multi-model AI applications. Integrate Ling 3.0 Flash without maintaining a separate provider-specific connection, and keep the flexibility to test, switch, or combine models as your application requirements evolve.

01

One API integration

Access Ling 3.0 Flash and other AI models through the same API workflow instead of maintaining separate integrations for every provider.

02

Easy model switching

Test Ling 3.0 Flash against alternative models or switch models as your performance, capability, or cost requirements change without rebuilding your application around another provider API.

03

Flexible for production

Use Ling 3.0 Flash from experimentation through production while keeping your AI stack flexible as workloads, traffic, and model requirements evolve.

04

Multi-model applications

Use Ling 3.0 Flash for the workloads where it performs best and combine it with other models for tasks that require different capabilities, performance, or efficiency.

Start free
arrow black icon
Read Docs

Frequently Asked Questions

Answers to common questions about integrating and using this AI model via AnyAPI.ai

Ling 3.0 Flash Sante is inclusionAI's health and medicine fine-tune of Ling 3.0 Flash, aimed at medical knowledge reasoning, clinical safety, evidence-based retrieval and long-horizon medical tasks. It suits medical literature search, clinical decision-support prototypes and health-focused assistants. It is a developer API for research and workflow assistance, not a medical device, and outputs should be reviewed by qualified professionals.

Ling 3.0 Flash Sante supports a 262,144-token context window and up to 32,768 output tokens per request. It inherits this large context from the base Ling 3.0 Flash, allowing multi-document medical reasoning across guidelines, literature and records in one call.

No. As of its release there is no independent third-party benchmark, Artificial Analysis page, or arena result for this exact fine-tune. The medical scores it reports — such as 53.88 on MedXpertQA-Text and 83.83 on DiagnosisArena-MCQ — are self-reported by inclusionAI and should be treated as vendor claims pending independent verification.

No. Unlike the open-weight base Ling 3.0 Flash, Sante is a proprietary API checkpoint whose weights were not released on Hugging Face. It can only be used through hosted API access, which means it cannot be self-hosted, fine-tuned further, or independently audited from weights.

Ling 3.0 Flash Sante accepts tools and tool_choice for function calling, making it usable in agent-style medical workflows. However, it does not support response_format, so JSON output is not schema-enforced. Applications needing strict structured responses should add their own validation layer.

* Benchmark data source: Artificial Analysis artificialanalysis.ai