DeplAI
— credits
DeplAI

Application

Dashboard

HomeYour ProfileOrganizationsUsageDocumentation

Services

UI/UX customizerSecurity AgentDASTCloudDeployInstance ManagementCode ReviewerSoonSessions

BYOK

KeysCatalogCompareUsage

Account

BillingInvoicesCreditsRefer & EarnNEWIntegrations
Settings
DE

DeplAI

Free

— creditsFree
deplaiDocumentation

Start here

Services

Account and models

Help and reference

Account and models

BYOK model catalog (September 2026)

This page mirrors the **Compare** view at `/dashboard/ai/compare`. Catalog refreshed **2026-09-01**.

This page mirrors the Compare view at /dashboard/ai/compare. Catalog refreshed 2026-09-01.

Best performance

For peak DeplAI performance—security analysis, deploy pipelines, and multi-step agents—use:

ModelProvider idWhen to use
MiniMax M3MiniMax-M3Agentic reasoning, coding, and 1M multimodal context. Set high or extrahigh thinking effort.
Grok 4.6grok-4.6Flagship xAI model for coding and long-running agents. Set high or extrahigh (xhigh) reasoning effort.

Both models expose a shared effort ladder from low through extrahigh. Reasoning cannot be disabled on Grok 4.6; MiniMax M3 defaults to adaptive thinking.

Platform baseline

DeplAI uses GPT-5.6 Sol (gpt-5.6-sol) as the internal platform baseline for routing and cost accounting. Compare rankings are relative to DeplAI jobs—not a generic chat leaderboard.

Active flagship models

Display nameProviderAPI idContextInput / output (USD per 1M tokens)
MiniMax M3minimaxMiniMax-M31M$0.60 / $2.40
Grok 4.6xaigrok-4.6500k$2.00 / $6.00
Grok 4.5xaigrok-4.5500k$2.00 / $6.00
Gemini 3.1 Progeminigemini-3.1-pro1M$2.00 / $12.00
GPT-5.6 Solopenaigpt-5.6-sol1.05M$4.00 / $20.00
GPT-5.6 Terraopenaigpt-5.6-terra1.05M$2.00 / $12.00
Claude Fable 5anthropicclaude-fable-51M$10.00 / $50.00
Claude Opus 5anthropicclaude-opus-51M$5.00 / $25.00
Claude Sonnet 5anthropicclaude-sonnet-51M$3.00 / $15.00

Other active models

Display nameProviderAPI idNotes
Gemini 3.5 Flash-Litegeminigemini-3.5-flash-liteFast, low-cost Gemini tier
MiniMax M2.7minimaxMiniMax-M2.7Prior M-series; always-on reasoning
Kimi K2.5kimikimi-k2.52M context
GLM-5glmglm-5Strong coding and agents
GPT-5.6 Lunaopenaigpt-5.6-lunaFast OpenAI tier
GPT-5.4 Miniopenaigpt-5.4-miniCost-optimized
GPT-5.3 Codexopenaigpt-5.3-codexCoding specialist
Claude Haiku 4.5anthropicclaude-haiku-4-5Fast Anthropic tier
Llama 3.3 70B (Groq)groqllama-3.3-70b-versatileHosted inference
GPT OSS 120B (Groq)groqopenai/gpt-oss-120bOpen-weight on Groq

Deprecated (still visible, not selectable)

ModelReplacement
Gemini 2.5 ProGemini 3.1 Pro
Gemini 2.5 FlashGemini 3.5 Flash-Lite
Claude Opus 4.6Claude Opus 5
Claude Sonnet 4.6Claude Sonnet 5
Grok 4Grok 4.6

Thinking effort by provider

ProviderParameterShared ladderDefault
OpenAIreasoning.effortlow · medium · high · extrahighmedium
Anthropicoutput_config.effortlow · medium · high · extrahighhigh
xAIreasoning_effortlow · medium · high · extrahighhigh
Geminithinking_levellow · medium · highmedium
MiniMaxthinking.type (M3)low · medium · high · extrahighmedium

OpenAI and Anthropic also expose extended API values (max, none, product-only ultra). xAI maps extrahigh to API xhigh.

Logical aliases

Agents can request logical names instead of a provider id:

AliasResolves to capability
best / best_reasoningHighest reasoning score among your keyed providers
best_codingStrongest coding model
best_agentBest agentic / tool-use fit
best_fastFast latency profile
best_costLowest output price
best_long_contextLargest context window
best_multimodalVision + audio where supported

With keys for MiniMax and xAI, best_reasoning typically ranks MiniMax M3 or Grok 4.6 first—still use high or extrahigh effort for peak results.

Related: Core concepts · Security and data · Billing

On this page

Best performancePlatform baselineActive flagship modelsOther active modelsDeprecated (still visible, not selectable)Thinking effort by providerLogical aliases