Ling 3.0 Flash reaches Vercel AI Gateway with a free access window

Vercel added Ling 3.0 Flash to AI Gateway on July 23, 2026, giving developers a short free access window for a new Ant Group model aimed at high-frequency agent workloads. The confirmed availability is through Vercel's gateway and OpenRouter's public model page, so treat this as a distribution and access update rather than proof that every vendor channel has published a full standalone launch page.

The practical detail is the deadline. Vercel says Ling 3.0 Flash is free to use for the next three weeks, through August 3. That makes the model worth testing quickly if your stack already uses AI SDK, OpenAI-compatible routing, or gateway-based provider failover.

Ling 3.0 Flash enters the gateway layer

Vercel lists Ling 3.0 Flash as an AI Gateway model under the slug inclusionai/ling-3.0-flash-free. The changelog describes it as a Mixture-of-Experts model with 124B total parameters, about 5.1B active parameters per token, a 256K token context window, and both thinking and non-thinking modes.

That profile is tuned for agent loops rather than one-off chat novelty. Vercel's positioning is specific: coding agents, document work, long-context multi-turn interactions, and repeated agent runs where token use, latency, and serving cost matter. OpenRouter's model page corroborates the same model family and lists Ling 3.0 Flash with 124B total parameters, about 5.1B active parameters per token, and a 262,144-token context window.

The free window is the reason to test now

Most model additions can wait until a team has a benchmark slot open. This one has a clock. Vercel says the free period runs through August 3, while its model page shows the provider as Novita AI and marks the release date as July 23, 2026.

For teams already using Vercel AI Gateway, the first test is simple: route a narrow class of agent tasks to inclusionai/ling-3.0-flash-free and compare output quality, latency, retries, and token count against your current low-cost model. Good candidates are issue triage, repository summarization, long support-ticket clustering, structured document extraction, and multi-step research notes where a 256K context window can reduce chunking overhead.

The caution is just as important. Free gateway availability is not the same as a permanent price cut, an open-weight release, or a universal API entitlement. Vercel says AI Gateway reflects provider pricing with no markup and no platform fee, including BYOK requests, but the free model entry should be treated as time-limited until the provider pages say otherwise.

What to verify in your own stack

Developers should check three things before moving real workload volume. First, confirm whether your compliance posture allows the listed provider route. Vercel's AI Gateway supports routing controls such as Zero Data Retention selection and provider allowlists, but the model page still points to a specific provider path that your team may need to approve.

Second, test both thinking and non-thinking behavior on your own prompts. MoE models can look excellent on short demos and still behave differently under long agent traces, strict JSON, or tool-heavy workflows. The model supports tool parameters through OpenRouter and AI Gateway, but support is only useful if your orchestration layer handles retries and malformed calls cleanly.

Third, measure cost after the trial. A free three-week window is useful for evaluation, not a pricing guarantee. Save baseline traces now: token count, wall-clock latency, answer acceptance rate, and fallback frequency. That will tell you whether Ling 3.0 Flash belongs in the fast lane, a cheap background-worker lane, or only as a temporary test route.

Source check

The primary source is Vercel's July 23 AI Gateway changelog, which gives the model name, Ant Group attribution, free-use window, parameter profile, 256K context claim, and AI SDK slug. Vercel's live model page confirms the gateway listing, provider route, release date, and model-card positioning. OpenRouter's public model page independently shows Ling 3.0 Flash with matching MoE scale and context-window details.

InclusionAI's own release archive confirms the broader Ling/Ming model-family context but did not surface a separate Ling 3.0 Flash launch page during this quick scan. That is why the safe reading is straightforward: Ling 3.0 Flash is now available through major routing gateways, with a time-limited free test window, while teams should still verify provider terms and long-term pricing before standardizing on it.

For a broader routing and agent-tooling checklist, keep LinkLoot's AI agent guide handy: /guides/ai-agent-tools.