Esta página está disponível em inglês. Sua preferência por português foi mantida para as páginas traduzidas.
/Models · inference and detection
Explore MiniMax M3, DeepSeek V4.1 Flash, Qwen 3.8, GLM 5.3, Nemotron, Inkling and Kimi K3. Shortlist up to three models and compare the same workload before you plan an evaluation.
Access is by invitation. Your account model list confirms what you can call. Models marked By request need deployment scoping; Soon models are awaiting launch.
8 of 21 models · Highlights + your selections · 0 of 3 shortlisted
Multimodal chat and agents
USD / 1M · Input $0.28 · Cached input $0.056 · Output $1.10
Published capabilities: Chat · Images · Video · Tools · Reasoning
License: MiniMax Community
Publisher model cardReasoning, coding and agents
USD / 1M · Input $0.17 · Cached input $0.169 · Output $3.00
Published capabilities: Chat · Tools · Reasoning
License: GLM 5.3 License
Publisher model cardMultimodal agents and coding
USD / 1M · Input $0.044 · Cached input $0.04 · Output $0.60
Published capabilities: Chat · Images · Video · Tools · Reasoning
License: MIT
Publisher model cardChat, tools and reasoning
USD / 1M · Input $0.0893 · Cached input $0.085 · Output $2.35
Published capabilities: Chat · Images · Video · Tools · Reasoning
License: Apache 2.0
Publisher model cardMulti-agent workflows
Deployment and account access require scoping.
USD / 1M · Input $0.085 · Cached input $0.0425 · Output $0.40
Published capabilities: Chat · Tools · Reasoning
License: NVIDIA Nemotron Open Model License
Publisher model cardChat, tools and reasoning
Rates confirmed for your account and configuration.
Published capabilities: Chat · Tools · Reasoning
Proposed S Series targets. Not Artificial Analysis scores.
Chat, tools and reasoning
Rates confirmed for your account and configuration.
Published capabilities: Chat · Tools · Reasoning
Proposed S Series targets. Not Artificial Analysis scores.
Image understanding, coding and agents
Deployment and account access require scoping.
USD / 1M · Input $0.09 · Cached input $0.018 · Output $0.18
Published capabilities: Chat · Images · Tools · Reasoning
License: MIT
Publisher model cardCapabilities and published context describe the publisher model or checkpoint. Your Grid deployment confirms enabled modalities, context and precision. Token rates are indicative new-account references; existing agreements retain their effective rates. See sources and pricing basis.
Intelligence is the Artificial Analysis Intelligence Index, v4.1.1. Reliability is the AA-Omniscience Index. It rewards correct answers, penalizes guesses and ignores abstention. Snapshot 16 August 2026. Kimi K3 uses max-effort scores. Unscored models show no marker. LIVIA S2.6 and S2.6 Flash chips are proposed S Series targets.
Specialist models / Detection
Choose AI-text review, prompt protection or both. Evaluation is by request. The pilot confirms the served version, coverage, account access and price.
Prioritize text for human review using AI-generation signals.
A detection signal supports review. It does not prove authorship or justify an automatic adverse decision.
Evaluate prompt injection and jailbreak signals before content reaches a model or tool.
Detection complements access controls, tool permissions and review. A low score does not establish that content is safe.
/your-models
Register a model you own and it joins your catalog. It is callable by id like any other entry here, and Smart Router can route to it if you want it in the mix. Teams that need isolation or a clear data boundary usually run these on Grid Private.
Register an adapter over a model that is already here and it becomes routable the same way. Qwen 3.8 27B is the adapter starting point. Explore other published checkpoints through custom training scoping.
/smart-router
Send model: grid-auto instead of an id. Routing is off unless you ask for it, request by request, so it is a mode and never a default.
Smart Router chooses among the models enabled for your account. It picks a lane capable of the request. Models listed for scoping join the routing pool after deployment and account access are confirmed.
Every routed response carries x-grid-routed-model. That header names the model that actually answered, so you can log it beside your own request id and reconstruct the call later.
Send an id and that id is what runs. A pinned version does not move under you when the catalog changes, which is what you want for determinism and release control.
Connect is Grid inside the coding agents and IDEs your engineers already use. Smart Router is still the engine. Invitation, not a second factory.
'/%3e%3cpath%20id='Vector_2'%20d='M8.91868%2014.1714H6.94748L5.84306%2012.3872H2.15192L3.28361%2010.603H5.1769L1.05919%203.67847H3.28361L5.43012%200.147056L6.42936%201.93127L5.43012%203.67847H13.0539L12.0547%205.35847L13.948%208.64446H12.0547L11.0652%206.89726L7.17148%2014.1714H8.91868Z'%20fill='white'/%3e%3cpath%20id='Vector_3'%20d='M9.76963%206.06834H4.73255L7.17123%2010.186L9.76963%206.06834Z'%20fill='url(%23paint1_radial_0_85)'/%3e%3c/g%3e%3cdefs%3e%3cradialGradient%20id='paint0_radial_0_85'%20cx='0'%20cy='0'%20r='1'%20gradientUnits='userSpaceOnUse'%20gradientTransform='translate(7.0565%208.03285)%20rotate(90)%20scale(9.73914%209.73914)'%3e%3cstop%20stop-color='%23665CEE'/%3e%3cstop%20offset='1'%20stop-color='%23332E91'/%3e%3c/radialGradient%3e%3cradialGradient%20id='paint1_radial_0_85'%20cx='0'%20cy='0'%20r='1'%20gradientUnits='userSpaceOnUse'%20gradientTransform='translate(7.05631%208.03272)%20rotate(90)%20scale(9.73914%209.73914)'%3e%3cstop%20stop-color='%23665CEE'/%3e%3cstop%20offset='1'%20stop-color='%23332E91'/%3e%3c/radialGradient%3e%3c/defs%3e%3c/svg%3e)
