Frontier Models

InclusionAI Launches Ling 3.0 Flash Fin Finance Variant

InclusionAI has released a finance-specialized variant of its Ling 3.0 Flash model that improves benchmark performance and offers free access periods before open-sourcing the weights.

By Marcus Vance August 28, 2026 2 MIN READ

Inside a modern open-plan fintech office headquarters associated with a major Chinese financial technology group several anonymous professionals in business casual attire gather around a large central conference table covered with printed financial reports stacks of bound documents and physical calculators while multiple computer workstations display abstract multicolored charts and numerical grids on their screens one analyst seated at the head of the table gestures toward a nearby rack of humming server hardware cabinets with visible cabling and indicator lights another professional stands beside a whiteboard covered in handwritten diagrams of data flows and model architectures while a third individual reviews printed benchmark comparison sheets spread across the table surface the room features floor-to-ceiling windows revealing an urban skyline with distant skyscrapers rows of identical ergonomic office chairs line the perimeter potted plants sit in corners next to filing cabinets filled with labeled folders coffee mugs and notepads rest on side tables alongside wireless mice and external hard drives a fourth person leans over a laptop keyboard typing while colleagues in the background examine a large wall-mounted display unit showing complex line graphs and pie charts the overall environment includes subtle details such as wall posters with abstract geometric patterns representing algorithmic concepts scattered USB drives on desks and a distant view into an adjacent server room with blinking network equipment all elements grounded in a realistic corporate setting focused on finance technology development and deployment without any visible lettering logos or markings the scene captures a moment of collaborative review and discussion among the team members dressed in neutral colored shirts slacks and blouses with varied hairstyles and postures indicating concentration and engagement the physical objects include metallic server racks with ventilation grilles tangled ethernet cables on the floor ergonomic keyboard trays under desks and reflective glass surfaces on the tabletops creating a dense layered composition of human activity and technological infrastructure directly tied to the release of a specialized financial artificial intelligence model variant by the originating organization and its associated partners in the sector. — Illustration: AI Intel Report

Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI built on Ling 3.0 Flash with 5.1B active parameters out of 124B total.

InclusionAI, affiliated with Ant Group, introduced Ling 3.0 Flash Fin on August 27, 2026, as a specialized model for finance applications.

What are the main technical specifications?

The model keeps the architecture of Ling 3.0 Flash, featuring 124B total parameters and 5.1B active parameters according to OpenRouter.

It offers a 262K-token context window and supports up to 32K output tokens, along with reasoning and function calling capabilities.

Where is Ling 3.0 Flash Fin currently available?

Vercel has made it available on its AI Gateway for free through September 25, as stated in their changelog.

OpenRouter also provides free access for one month following the release.

Ling 3.0 Flash Fin is a finance-focused mixture-of-experts model from InclusionAI, built on Ling 3.0 Flash with 5.1B active parameters out of 124B total. It is designed for real-world investment workflows that require complex multi-step tasks and long-horizon planning and execution, while retaining general capabilities in reasoning, coding, and mathematics.InclusionAI

What benchmarks were used to evaluate the model?

It was tested on finance agent benchmarks including FinFIRST in collaboration with CICC, FinSearchComp Verified, Finance Agent, APEX-Agents, SpreadsheetBench, and τ³-Banking.

Selected finance benchmarks for Ling 3.0 Flash Fin evaluation
Benchmark	Focus Area
FinFIRST	Financial tasks with CICC
FinSearchComp Verified	Search and comprehension
SpreadsheetBench	Data handling in spreadsheets
τ³-Banking	Banking operations

Review performance gains on specialized tasks
Assess retention of general reasoning abilities
Plan for integration into investment workflows
Monitor open weights release for community use

What does this mean for stakeholders in AI and finance?

The release underscores efforts to create domain-adapted models that deliver targeted improvements without sacrificing base efficiency.

With open weights expected soon, financial institutions and developers may explore custom applications.

What developments are anticipated next?

The open-sourcing of weights next week will enable wider testing and potential fine-tuning by the AI community, according to IT之家.

Frequently asked

What is the context window size of Ling 3.0 Flash Fin?

The model supports a 262K-token context window. It also allows up to 32K output tokens.

Sources

OpenRouter — Describes the model as finance-focused MoE with parameter counts and provides the model description quote.
IT之家 — Reports the AA Intelligence Index score of 41 and the plan to open source weights next week.
Vercel — Announces free availability on AI Gateway through September 25.
AI News — Ant Group has recently partnered with China International Capital Corporation (CICC) ... Ling-3.0-flash-Fin, the first financial-enhanced large model under Ant Bailing. ... The model weights are also planned to be…