Skip to main content

Available Models

This page lists all models available through the PwC AI-CoE Coding Agents Gateway: the models of the PwC GenAI Shared Service plus a set of self-hosted and partner models served via the PwC Agent orchestrator.

These models are safe to use as they run on dedicated instances provisioned for PwC by PwC Global. They are the same models powering the IDI (Instant Document Interaction) platform. For more information, visit the official PwC GenAI Shared Service SharePoint.

Model Compatibility

Different coding agents work best with different model families:

Coding AgentCompatible ModelsRecommendation
Claude CodeAnthropic models (Claude Opus, Sonnet, Haiku), self-hosted models, and autoAnthropic models recommended
ClineAll supported modelsOptimized for Anthropic models
CodexOpenAI models with Responses API support onlySee Responses API column below
General Guidance

Use models from the same vendor as your coding agent for the best experience. For example, use Claude models with Claude Code, and OpenAI models with Codex.

Supported Models

Configuration Note

When configuring your coding agent, you can use any of the three model ID formats accepted by the gateway:

  • Vendor naming convention (e.g., claude-sonnet-4-6, gpt-5.2-codex) — shortest, routes to the default provider for that model
  • PwC GenAI Shared Service naming convention (e.g., bedrock.anthropic.claude-opus-4-6) — provider-explicit shorthand
  • Coding Agents Gateway naming convention (e.g., GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_46_OPUS) — recommended when you want to pin a specific cloud provider

The Vendor Name column below shows the bare name for each model. Where two rows share the same model (different providers), only the default-provider row shows a bare name — use the GENAI internal name to target the other provider explicitly.

Models marked with ⭐ are our recommended choices for the best balance of performance and cost.

Model IDs

The gateway accepts three formats for the model field (or model-name custom header):

FormatExampleNotes
Vendor naming conventionclaude-sonnet-4-6Shortest. Routes to the default provider. Date suffixes are stripped automatically (e.g., claude-sonnet-4-5-20241022claude-sonnet-4-5).
PwC GenAI Shared Service naming conventionbedrock.anthropic.claude-opus-4-6Provider-explicit shorthand.
Coding Agents Gateway naming conventionGENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_46_OPUSUnambiguous. Recommended when pinning a specific cloud provider.

Anthropic Models

Official NameVendor NameGateway Model ID 🔑PwC Cloud ProviderInput ($/1M)Output ($/1M)ContextMax OutputResponses API
Claude Opus 4.8claude-opus-4-8GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_48_OPUSVertex AI$5.50$27.501M128k-
Claude Opus 4.8-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_48_OPUSBedrock$5.50$27.501M128k-
Claude Opus 4.7claude-opus-4-7GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_47_OPUSVertex AI$5.50$27.501M128k-
Claude Opus 4.7-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_47_OPUSBedrock$5.50$27.501M128k-
Claude Opus 4.6claude-opus-4-6GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_46_OPUSVertex AI$5.50$27.501M128k-
Claude Opus 4.6-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_46_OPUSBedrock$5.50$27.501M128k-
Claude Opus 4.5-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_45_OPUSBedrock$5.00$25.00200k64k-
Claude Opus 4.5claude-opus-4-5GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_45_OPUSVertex AI$5.00$25.00200k64k-
Claude Sonnet 5claude-sonnet-5GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_5_SONNETVertex AI$2.20$11.001M128k-
Claude Sonnet 4.6claude-sonnet-4-6GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_46_SONNETVertex AI$3.30$16.501M64k-
Claude Sonnet 4.6-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_46_SONNETBedrock$3.30$16.501M64k-
Claude Sonnet 4.5claude-sonnet-4-5GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_45_SONNETVertex AI$3.00$15.00200k64k-
Claude Sonnet 4.5-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_45_SONNETBedrock$3.00$15.00200k64k-
Claude Sonnet 4claude-sonnet-4GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_4_SONNETVertex AI$3.00$15.001M64k-
Claude Sonnet 4-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_4_SONNETBedrock$3.00$15.001M64k-
Claude Haiku 4.5claude-haiku-4-5GENAI_SHARED_VERTEXAI_ANTHROPIC_CLAUDE_45_HAIKUVertex AI$1.00$5.00200k8k-
Claude Haiku 4.5-GENAI_SHARED_BEDROCK_ANTHROPIC_CLAUDE_45_HAIKUBedrock$1.00$5.00200k8k-
1M Context

Models marked with 1M context (Opus 4.8, Opus 4.7, Opus 4.6, Sonnet 5, Sonnet 4.6, Sonnet 4) support extended context via the anthropic-beta: context-1m-2025-08-07 header. The gateway automatically injects this header — no client configuration needed. The full 1M context window is included at standard pricing with no extra cost.

OpenAI Models

Official NameVendor NameGateway Model ID 🔑PwC Cloud ProviderInput ($/1M)Output ($/1M)ContextMax OutputResponses API
GPT-5.2 Codexgpt-5.2-codexGENAI_SHARED_OPENAI_GPT_52_CODEXOpenAI$1.75$14.00400k128k
GPT-5.2gpt-5.2GENAI_SHARED_OPENAI_GPT_52OpenAI$1.75$14.00400k128k-
GPT-5.1 Codexgpt-5.1-codexGENAI_SHARED_OPENAI_GPT_51_CODEXOpenAI$1.25$10.00272k128k
GPT-5.1gpt-5.1GENAI_SHARED_OPENAI_GPT_51OpenAI$1.25$10.00272k128k-
GPT-5 Codexgpt-5-codexGENAI_SHARED_OPENAI_GPT_5_CODEXOpenAI$1.25$10.00272k128k
GPT-5.4gpt-5.4GENAI_SHARED_OPENAI_OPENAI_GPT_54OpenAI$2.50$15.001M128k
GPT-5.4 Minigpt-5.4-miniGENAI_SHARED_OPENAI_OPENAI_GPT_54_MINIOpenAI$0.75$4.50400k128k
GPT-5.4 Nanogpt-5.4-nanoGENAI_SHARED_OPENAI_OPENAI_GPT_54_NANOOpenAI$0.20$1.25400k128k
GPT-5.4 Progpt-5.4-proGENAI_SHARED_OPENAI_OPENAI_GPT_54_PROOpenAI--400k128k
GPT-5.5-GENAI_SHARED_OPENAI_OPENAI_GPT_55OpenAI$5.00$30.001M128k
GPT-5.5gpt-5.5GENAI_SHARED_AZURE_OPENAI_GPT_55Azure$5.00$30.001M128k
GPT-5.6 Luna (EU)gpt-5.6-lunaGENAI_SHARED_OPENAI_OPENAI_GPT_56_LUNAOpenAI$1.00$6.001M128k-
GPT-5.6 Sol (EU)gpt-5.6-solGENAI_SHARED_OPENAI_OPENAI_GPT_56_SOLOpenAI$5.00$30.001M128k-
GPT-5.6 Terra (EU)gpt-5.6-terraGENAI_SHARED_OPENAI_OPENAI_GPT_56_TERRAOpenAI$2.50$15.001M128k-
GPT-5gpt-5GENAI_SHARED_AZURE_OPENAI_GPT_5Azure$1.25$10.00272k128k-
GPT-5 Minigpt-5-miniGENAI_SHARED_AZURE_OPENAI_GPT_5_MINIAzure$0.25$2.00272k128k-
GPT-5 Nanogpt-5-nanoGENAI_SHARED_AZURE_OPENAI_GPT_5_NANOAzure$0.05$0.40272k128k-
Responses API

The Responses API column indicates which models support OpenAI's Responses API format. This is required for Codex coding agent compatibility. Models marked with ✅ can be used with Codex; models marked with - cannot.

Google Models

Official NameVendor NameGateway Model ID 🔑PwC Cloud ProviderInput ($/1M)Output ($/1M)ContextMax OutputResponses API
Gemini 2.5 Progemini-2.5-proGENAI_SHARED_VERTEXAI_GOOGLE_GEMINI_25_PROVertex AI$1.25$10.001M65k-
Gemini 2.5 Flashgemini-2.5-flashGENAI_SHARED_VERTEXAI_GOOGLE_GEMINI_25_FLASHVertex AI$0.30$2.501M65k-
Gemini 2.5 Flash Litegemini-2.5-flash-liteGENAI_SHARED_VERTEXAI_GOOGLE_GEMINI_25_FLASH_LITEVertex AI$0.10$0.401M65k-

Self-Hosted and Partner Models (PwC Agent)

These models are served through the PwC Agent orchestrator (self-hosted on PwC infrastructure or via Azure AI Foundry). Use the Model ID shown as the model value.

Official NameModel IDHostingNotes
GLM-5.2glmSelf-hostedGeneral coding model
Qwen3-27B (thinking)qwen-thinkingSelf-hostedThinking mode enabled
Qwen3-27B (fast)qwen-fastSelf-hostedThinking mode disabled
DeepSeek V4 Flash (thinking)ds4Self-hostedThinking mode enabled
DeepSeek V4 Flash (fast)ds4-fastSelf-hostedThinking mode disabled
Ternary-Bonsai-27BbonsaiSelf-hostedFallback model
Claude Fable 5claude-fable-5Vertex AI (PwC Agent)
DeepSeek V4 ProDeepSeek-V4-ProAzure AI Foundry
Kimi K2.6Kimi-K2.6Azure AI Foundry

Discovering your models (/models command)

GET /models (OpenAI shape) and GET /v1/models (Anthropic shape) return the models available to your API key — restricted models (e.g. Claude Fable 5, GPT-5.6 Sol) are listed only if your key is entitled.

Claude Code users can install a /models slash command that shows this list directly in the CLI:

mkdir -p ~/.claude/commands
curl -fsSL https://idi-coding-agents.pwc.it/downloads/models-command.md -o ~/.claude/commands/models.md

Then type /models inside Claude Code to get a table of the model IDs you can use with /model <id>.

Automatic Model Selection (auto)

Set model to auto to let the gateway pick the best model for each request. A task classifier analyses your message (task type and difficulty) and routes it to an appropriate backend — lightweight self-hosted models for simple tasks, Claude Sonnet/Opus for hard coding or planning tasks.

Embedding Models

Available via the /embeddings endpoint:

Official NameVendor NameGateway Model ID 🔑PwC Cloud Provider
Text Embedding 3 Largetext-embedding-3-largeGENAI_SHARED_AZURE_OPENAI_TEXT_EMBEDDING_003_LARGEAzure
Text Embedding Ada 002text-embedding-ada-002GENAI_SHARED_AZURE_OPENAI_TEXT_ADA_002Azure
Titan Embed Image v1titan-embed-image-v1GENAI_SHARED_BEDROCK_AMAZON_TITAN_EMBED_IMAGE_V1Bedrock
Gemini Embeddinggemini-embeddingGENAI_SHARED_VERTEXAI_GEMINI_EMBEDDINGVertex AI