Function getModelMaxOutputTokens

src/utils/context.ts:197–258

(model: string)

 * Returns the model's default and upper limit for max output tokens.
 */
export function getModelMaxOutputTokens(model: string): {
  default: number
  upperLimit: number
} {
  let defaultTokens: number
  let upperLimit: number

  if (process.env.USER_TYPE === 'ant') {
    const antModel = resolveAntModel(model.toLowerCase())
    if (antModel) {
      defaultTokens = antModel.defaultMaxTokens ?? MAX_OUTPUT_TOKENS_DEFAULT
      upperLimit = antModel.upperMaxTokensLimit ?? MAX_OUTPUT_TOKENS_UPPER_LIMIT
      return { default: defaultTokens, upperLimit }
    }
  }

  const m = getCanonicalName(model)

  if (m.includes('opus-4-6')) {
    defaultTokens = 64_000
    upperLimit = 128_000
  } else if (m.includes('sonnet-4-6')) {
    defaultTokens = 32_000
    upperLimit = 128_000
  } else if (
    m.includes('opus-4-5') ||
    m.includes('sonnet-4') ||
    m.includes('haiku-4')
  ) {
    defaultTokens = 32_000
    upperLimit = 64_000
  } else if (m.includes('opus-4-1') || m.includes('opus-4')) {
    defaultTokens = 32_000
    upperLimit = 32_000
  } else if (m.includes('claude-3-opus')) {
    defaultTokens = 4_096
    upperLimit = 4_096
  } else if (m.includes('claude-3-sonnet')) {
    defaultTokens = 8_192
    upperLimit = 8_192
  } else if (m.includes('claude-3-haiku')) {
    defaultTokens = 4_096
    upperLimit = 4_096
  } else if (m.includes('3-5-sonnet') || m.includes('3-5-haiku')) {
    defaultTokens = 8_192
    upperLimit = 8_192
  } else if (m.includes('3-7-sonnet')) {
    defaultTokens = 32_000
    upperLimit = 64_000
  } else {
    defaultTokens = MAX_OUTPUT_TOKENS_DEFAULT
    upperLimit = MAX_OUTPUT_TOKENS_UPPER_LIMIT
  }

  const cap = getModelCapability(model)
  if (cap?.max_tokens && cap.max_tokens >= 4_096) {
    upperLimit = cap.max_tokens
    defaultTokens = Math.min(defaultTokens, upperLimit)