[snehmehta]

Description

Adds pricing and model configs for Frequency (frequency), a voice-first, India-first, low-latency inference provider. Inference runs in India, behind an OpenAI-compatible API at https://llm.scalefrequency.com/v2.

Disclosure: I work on Frequency, so this is the provider adding itself.

Companion gateway PR: Portkey-AI/gateway#1872

• New model pricing
• Update existing pricing
• New model configuration
• Bug fix

Files

• pricing/frequency.json: pay-as-you-go rates. The default entry has zero prices and a calculate block that bills cache_read_tokens at the cached rate, in the same shape as novita-ai.json and moonshot.json.
• general/frequency.json: params, the max_tokens ceiling and capability flags.

Pricing

Prices are in cents per token (USD per 1M ÷ 10,000):

Model	Input $/1M	Cached input $/1M	Output $/1M
google/gemma-4-26b-a4b-it	0.15 → 0.000015	0.093 → 0.0000093	0.55 → 0.000055

mira-llm gets the same rates. It is the API's name for the same model, it is accepted as a request model, and the docs say responses report model: "mira-llm", so pricing it keeps cost tracking right whichever id a log carries. In general/ I listed only google/gemma-4-26b-a4b-it so the model picker doesn't show the same model twice.

Model config

Chat with tool calling. Text only, so no image flag. Context is 32,768 tokens. The values below come from the docs' parameter table and the limits for a key a user creates themselves (https://docs.miraiminds.co/v2/llm#limits):

• max_tokens: up to 8,192 (400 when unset). The API lowers anything larger to 8,192.
• temperature: 0 to 1.5 (default 0.3); top_p 0 to 1; top_k 0 to 1,000 (0 = off); frequency_penalty and presence_penalty -2 to 2.
• response_format: json_object and json_schema.
• A request takes up to 100 messages, 16,000 characters of message text and 16 tools.

Source Verification

Source Link: https://scalefrequency.com/pricing

Rates are also listed in the API docs: https://docs.miraiminds.co/v2/llm (limits: https://docs.miraiminds.co/v2/llm#limits)

Checklist

• I have validated the JSON using jq or an online validator
• I have verified that prices are in cents per token (not dollars)
• I have included the source link above
• I have signed the CLA (if first-time contributor)

Checks run locally: jq empty on both files, python3 scripts/check_duplicate_keys.py (no duplicates), prettier@2.8.8 --check and eslint --ext .json on both files, and node scripts/sync-to-gateway.js against a scratch directory (frequency.json passes the pricing schema check). I also converted each price back to USD per 1M to confirm the values.

Related Issue

None.

🤖 Generated with Claude Code