Skip to content

feat: add Claude 4.5 Sonnet model across all providers #8007

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
daniel-lxs wants to merge 6 commits into from
Closed

feat: add Claude 4.5 Sonnet model across all providers #8007

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
64c1dea
Add coding plan support to Z.ai provider
daniel-lxs Sep 15, 2025
40d6155
Remove kilocode_change comments
daniel-lxs Sep 15, 2025
8a961f2
feat: add Claude 4.5 Sonnet and Opus models across all providers
daniel-lxs Sep 15, 2025
049f239
refactor: remove Claude 4.5 Opus model references from providers and …
daniel-lxs Sep 29, 2025
5e01840
refactor: update Claude model identifiers to use new naming convention
daniel-lxs Sep 29, 2025
ad518b9
revert: remove ZAi provider changes from PR
daniel-lxs Sep 29, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
227 changes: 171 additions & 56 deletions packages/types/src/provider-settings.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,53 +27,126 @@ import {
internationalZAiModels,
} from "./providers/index.js"

/**
* constants
*/

export const DEFAULT_CONSECUTIVE_MISTAKE_LIMIT = 3

/**
* DynamicProvider
*
* Dynamic provider requires external API calls in order to get the model list.
*/

export const dynamicProviders = [
"openrouter",
"vercel-ai-gateway",
"huggingface",
"litellm",
"deepinfra",
"io-intelligence",
"requesty",
"unbound",
"glama",
] as const

export type DynamicProvider = (typeof dynamicProviders)[number]

export const isDynamicProvider = (key: string): key is DynamicProvider =>
dynamicProviders.includes(key as DynamicProvider)

/**
* LocalProvider
*
* Local providers require localhost API calls in order to get the model list.
*/

export const localProviders = ["ollama", "lmstudio"] as const

export type LocalProvider = (typeof localProviders)[number]

export const isLocalProvider = (key: string): key is LocalProvider => localProviders.includes(key as LocalProvider)

/**
* InternalProvider
*
* Internal providers require internal VSCode API calls in order to get the
* model list.
*/

export const internalProviders = ["vscode-lm"] as const

export type InternalProvider = (typeof internalProviders)[number]

export const isInternalProvider = (key: string): key is InternalProvider =>
internalProviders.includes(key as InternalProvider)

/**
* CustomProvider
*
* Custom providers are completely configurable within Roo Code settings.
*/

export const customProviders = ["openai"] as const

export type CustomProvider = (typeof customProviders)[number]

export const isCustomProvider = (key: string): key is CustomProvider => customProviders.includes(key as CustomProvider)

/**
* FauxProvider
*
* Faux providers do not make external inference calls and therefore do not have
* model lists.
*/

export const fauxProviders = ["fake-ai", "human-relay"] as const

export type FauxProvider = (typeof fauxProviders)[number]

export const isFauxProvider = (key: string): key is FauxProvider => fauxProviders.includes(key as FauxProvider)

/**
* ProviderName
*/

export const providerNames = [
...dynamicProviders,
...localProviders,
...internalProviders,
...customProviders,
...fauxProviders,
"anthropic",
"claude-code",
"glama",
"openrouter",
"bedrock",
"vertex",
"openai",
"ollama",
"vscode-lm",
"lmstudio",
"cerebras",
"chutes",
"claude-code",
"doubao",
"deepseek",
"featherless",
"fireworks",
"gemini",
"gemini-cli",
"openai-native",
"groq",
"mistral",
"moonshot",
"deepseek",
"deepinfra",
"doubao",
"openai-native",
"qwen-code",
"unbound",
"requesty",
"human-relay",
"fake-ai",
"xai",
"groq",
"chutes",
"litellm",
"huggingface",
"cerebras",
"roo",
"sambanova",
"vertex",
"xai",
"zai",
"fireworks",
"featherless",
"io-intelligence",
"roo",
"vercel-ai-gateway",
] as const

export const providerNamesSchema = z.enum(providerNames)

export type ProviderName = z.infer<typeof providerNamesSchema>

export const isProviderName = (key: unknown): key is ProviderName =>
typeof key === "string" && providerNames.includes(key as ProviderName)

/**
* ProviderSettingsEntry
*/
Expand All @@ -91,11 +164,6 @@ export type ProviderSettingsEntry = z.infer<typeof providerSettingsEntrySchema>
* ProviderSettings
*/

/**
* Default value for consecutive mistake limit
*/
export const DEFAULT_CONSECUTIVE_MISTAKE_LIMIT = 3

const baseProviderSettingsSchema = z.object({
includeMaxTokens: z.boolean().optional(),
diffEnabled: z.boolean().optional(),
Expand Down Expand Up @@ -124,7 +192,7 @@ const anthropicSchema = apiModelIdProviderModelSchema.extend({
apiKey: z.string().optional(),
anthropicBaseUrl: z.string().optional(),
anthropicUseAuthToken: z.boolean().optional(),
anthropicBeta1MContext: z.boolean().optional(), // Enable 'context-1m-2025-08-07' beta for 1M context window
anthropicBeta1MContext: z.boolean().optional(), // Enable 'context-1m-2025-08-07' beta for 1M context window.
})

const claudeCodeSchema = apiModelIdProviderModelSchema.extend({
Expand Down Expand Up @@ -160,7 +228,7 @@ const bedrockSchema = apiModelIdProviderModelSchema.extend({
awsModelContextWindow: z.number().optional(),
awsBedrockEndpointEnabled: z.boolean().optional(),
awsBedrockEndpoint: z.string().optional(),
awsBedrock1MContext: z.boolean().optional(), // Enable 'context-1m-2025-08-07' beta for 1M context window
awsBedrock1MContext: z.boolean().optional(), // Enable 'context-1m-2025-08-07' beta for 1M context window.
})

const vertexSchema = apiModelIdProviderModelSchema.extend({
Expand Down Expand Up @@ -190,6 +258,7 @@ const ollamaSchema = baseProviderSettingsSchema.extend({
ollamaModelId: z.string().optional(),
ollamaBaseUrl: z.string().optional(),
ollamaApiKey: z.string().optional(),
ollamaNumCtx: z.number().int().min(128).optional(),
})

const vsCodeLmSchema = baseProviderSettingsSchema.extend({
Expand Down Expand Up @@ -308,9 +377,13 @@ const sambaNovaSchema = apiModelIdProviderModelSchema.extend({
sambaNovaApiKey: z.string().optional(),
})

export const zaiApiLineSchema = z.enum(["international_coding", "international", "china_coding", "china"])

export type ZaiApiLine = z.infer<typeof zaiApiLineSchema>

const zaiSchema = apiModelIdProviderModelSchema.extend({
zaiApiKey: z.string().optional(),
zaiApiLine: z.union([z.literal("china"), z.literal("international")]).optional(),
zaiApiLine: zaiApiLineSchema.optional(),
})

const fireworksSchema = apiModelIdProviderModelSchema.extend({
Expand All @@ -331,7 +404,7 @@ const qwenCodeSchema = apiModelIdProviderModelSchema.extend({
})

const rooSchema = apiModelIdProviderModelSchema.extend({
// No additional fields needed - uses cloud authentication
// No additional fields needed - uses cloud authentication.
})

const vercelAiGatewaySchema = baseProviderSettingsSchema.extend({
Expand Down Expand Up @@ -436,7 +509,11 @@ export type ProviderSettingsWithId = z.infer<typeof providerSettingsWithIdSchema

export const PROVIDER_SETTINGS_KEYS = providerSettingsSchema.keyof().options

export const MODEL_ID_KEYS: Partial<keyof ProviderSettings>[] = [
/**
* ModelIdKey
*/

export const modelIdKeys = [
"apiModelId",
"glamaModelId",
"openRouterModelId",
Expand All @@ -451,13 +528,63 @@ export const MODEL_ID_KEYS: Partial<keyof ProviderSettings>[] = [
"ioIntelligenceModelId",
"vercelAiGatewayModelId",
"deepInfraModelId",
]
] as const satisfies readonly (keyof ProviderSettings)[]

export type ModelIdKey = (typeof modelIdKeys)[number]

export const getModelId = (settings: ProviderSettings): string | undefined => {
const modelIdKey = MODEL_ID_KEYS.find((key) => settings[key])
return modelIdKey ? (settings[modelIdKey] as string) : undefined
const modelIdKey = modelIdKeys.find((key) => settings[key])
return modelIdKey ? settings[modelIdKey] : undefined
}

/**
* TypicalProvider
*/

export type TypicalProvider = Exclude<ProviderName, InternalProvider | CustomProvider | FauxProvider>

export const isTypicalProvider = (key: unknown): key is TypicalProvider =>
isProviderName(key) && !isInternalProvider(key) && !isCustomProvider(key) && !isFauxProvider(key)

export const modelIdKeysByProvider: Record<TypicalProvider, ModelIdKey> = {
anthropic: "apiModelId",
"claude-code": "apiModelId",
glama: "glamaModelId",
openrouter: "openRouterModelId",
bedrock: "apiModelId",
vertex: "apiModelId",
"openai-native": "openAiModelId",
ollama: "ollamaModelId",
lmstudio: "lmStudioModelId",
gemini: "apiModelId",
"gemini-cli": "apiModelId",
mistral: "apiModelId",
moonshot: "apiModelId",
deepseek: "apiModelId",
deepinfra: "deepInfraModelId",
doubao: "apiModelId",
"qwen-code": "apiModelId",
unbound: "unboundModelId",
requesty: "requestyModelId",
xai: "apiModelId",
groq: "apiModelId",
chutes: "apiModelId",
litellm: "litellmModelId",
huggingface: "huggingFaceModelId",
cerebras: "apiModelId",
sambanova: "apiModelId",
zai: "apiModelId",
fireworks: "apiModelId",
featherless: "apiModelId",
"io-intelligence": "ioIntelligenceModelId",
roo: "apiModelId",
"vercel-ai-gateway": "vercelAiGatewayModelId",
}

/**
* ANTHROPIC_STYLE_PROVIDERS
*/

// Providers that use Anthropic-style API protocol.
export const ANTHROPIC_STYLE_PROVIDERS: ProviderName[] = ["anthropic", "claude-code", "bedrock"]

Expand All @@ -478,6 +605,10 @@ export const getApiProtocol = (provider: ProviderName | undefined, modelId?: str
return "openai"
}

/**
* MODELS_BY_PROVIDER
*/

export const MODELS_BY_PROVIDER: Record<
Exclude<ProviderName, "fake-ai" | "human-relay" | "gemini-cli" | "lmstudio" | "openai" | "ollama">,
{ id: ProviderName; label: string; models: string[] }
Expand Down Expand Up @@ -575,19 +706,3 @@ export const MODELS_BY_PROVIDER: Record<
deepinfra: { id: "deepinfra", label: "DeepInfra", models: [] },
"vercel-ai-gateway": { id: "vercel-ai-gateway", label: "Vercel AI Gateway", models: [] },
}

export const dynamicProviders = [
"glama",
"huggingface",
"litellm",
"openrouter",
"requesty",
"unbound",
"deepinfra",
"vercel-ai-gateway",
] as const satisfies readonly ProviderName[]

export type DynamicProvider = (typeof dynamicProviders)[number]

export const isDynamicProvider = (key: string): key is DynamicProvider =>
dynamicProviders.includes(key as DynamicProvider)
22 changes: 22 additions & 0 deletions packages/types/src/providers/anthropic.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,28 @@ export type AnthropicModelId = keyof typeof anthropicModels
export const anthropicDefaultModelId: AnthropicModelId = "claude-sonnet-4-20250514"

export const anthropicModels = {
"claude-4.5-sonnet": {
maxTokens: 64_000, // Overridden to 8k if `enableReasoningEffort` is false.
contextWindow: 200_000, // Default 200K, extendable to 1M with beta flag 'context-1m-2025-08-07'
supportsImages: true,
supportsComputerUse: true,
supportsPromptCache: true,
inputPrice: 3.0, // $3 per million input tokens (≤200K context)
outputPrice: 15.0, // $15 per million output tokens (≤200K context)
cacheWritesPrice: 3.75, // $3.75 per million tokens
cacheReadsPrice: 0.3, // $0.30 per million tokens
supportsReasoningBudget: true,
// Tiered pricing for extended context (requires beta flag 'context-1m-2025-08-07')
tiers: [
{
contextWindow: 1_000_000, // 1M tokens with beta flag
inputPrice: 6.0, // $6 per million input tokens (>200K context)
outputPrice: 22.5, // $22.50 per million output tokens (>200K context)
cacheWritesPrice: 7.5, // $7.50 per million tokens (>200K context)
cacheReadsPrice: 0.6, // $0.60 per million tokens (>200K context)
},
],
},
"claude-sonnet-4-20250514": {
maxTokens: 64_000, // Overridden to 8k if `enableReasoningEffort` is false.
contextWindow: 200_000, // Default 200K, extendable to 1M with beta flag 'context-1m-2025-08-07'
Expand Down
15 changes: 15 additions & 0 deletions packages/types/src/providers/bedrock.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,21 @@ export const bedrockModels = {
maxCachePoints: 1,
cachableFields: ["system"],
},
"anthropic.claude-4.5-sonnet-v1:0": {
maxTokens: 8192,
Copy link
Contributor

@roomote roomote bot Sep 15, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The maxTokens value for Claude 4.5 models is 8,192 here but 64,000/32,000 in the anthropic.ts file. Is this intentional due to Bedrock limitations, or should these values be aligned?

contextWindow: 200_000,
supportsImages: true,
supportsComputerUse: true,
supportsPromptCache: true,
supportsReasoningBudget: true,
inputPrice: 3.0,
outputPrice: 15.0,
cacheWritesPrice: 3.75,
cacheReadsPrice: 0.3,
minTokensPerCachePoint: 1024,
maxCachePoints: 4,
cachableFields: ["system", "messages", "tools"],
},
"anthropic.claude-sonnet-4-20250514-v1:0": {
maxTokens: 8192,
contextWindow: 200_000,
Expand Down
8 changes: 8 additions & 0 deletions packages/types/src/providers/claude-code.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,14 @@ export function getClaudeCodeModelId(baseModelId: ClaudeCodeModelId, useVertex =
}

export const claudeCodeModels = {
"claude-4.5-sonnet": {
...anthropicModels["claude-4.5-sonnet"],
supportsImages: false,
supportsPromptCache: true, // Claude Code does report cache tokens
supportsReasoningEffort: false,
supportsReasoningBudget: false,
requiredReasoningBudget: false,
},
"claude-sonnet-4-20250514": {
...anthropicModels["claude-sonnet-4-20250514"],
supportsImages: false,
Expand Down
Loading
Loading