AI News Pulse

Subscriptions · API only

API only

7 pay-per-token APIs, straight from the labs, from 7 providers, cheapest first. Each card says how the limits work, the rules that come with the plan, and the models it includes.

62 plans from 17 providers, checked

API only

7 entries, cheapest first

  • Alibaba Cloud (Model Studio / DashScope)

    Model Studio API (pay-as-you-go)

    API onlyPay per tokenClear quotaOff-peak 22:00-08:00 (night), for qwen3.7-plus on Global deployments only UTC+8: Night: 60% off (limited-time); daytime: 20% off (limited-time), as labelled on the page
    Pay per use
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    Qwen3.8 Max (0902)$2$6per 1M tokens, Cached input $0.25; cache write $2.50; API name qwen3.8-max
    Qwen3.7 Max$1.475$4.425per 1M tokens, Cached input $0.295; cache write $1.84375
    Qwen3.7 Plus$0.32$1.28per 1M tokens, Cached input $0.064; cache write $0.40
    Qwen3.8 Flash$0.15$0.47per 1M tokens, Cached input $0.016; cache write $0.20
    Qwen3.6 Flash$0.1875$1.125per 1M tokens, cache write $0.234375
    Qwen3.6 Plus$0.325$1.95per 1M tokens, cache write $0.40625

    Rates from OpenRouter, Alibaba's own endpoint, checked 2026-10-21

    Models
    • qwen3.8-max
    • qwen3.7-max
    • qwen3.7-plus
    • qwen3.8-flash
    • qwen3.6-flash
    Worth knowing
    Off-peak: 22:00-08:00 (night), for qwen3.7-plus on Global deployments only (UTC+8), Night: 60% off (limited-time); daytime: 20% off (limited-time), as labelled on the page.
    More details
    • Batch API: 50% discount listed for qwen3.8-max, qwen3.7-max and qwen3.8-flash (Beijing) and qwen3.6-flash (Singapore and Beijing); batch and context-cache discounts cannot be combined
    • Explicit cache creation is billed at 125% of the standard input price (page)
    • Rates are Alibaba's default endpoint on OpenRouter; OpenRouter also lists a us-east-1 endpoint, priced higher for Qwen3.7 Max. The API name qwen3.8-max is served as Qwen3.8 Max (0902).
    Verified Source ↗Model pricing (Chinese page, CNY) ↗Get an API key
  • DeepSeek

    DeepSeek API (pay-as-you-go)

    API onlyPay per tokenClear quotaOff-peak All hours except peak: peak is 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, excluding Chinese public holidays. Weekends and Chinese public holidays are off-peak all day UTC (Chinese page: peak is 09:00-12:00 and 14:00-18:00 Beijing time, weekdays): 50% (off-peak rates are half of peak rates)
    No subscription
    API only
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    DeepSeek V4.1 Flash 1017$0.15$0.60per 1M tokens, Cached input $0.003; API name deepseek-flash; read in DeepSeek's off-peak hours
    DeepSeek V4 Pro 0813$0.66$1.98per 1M tokens, Cached input $0.022; API name deepseek-v4-pro; read in DeepSeek's off-peak hours

    Rates from OpenRouter, DeepSeek's own endpoint, checked 2026-10-21

    Models
    • deepseek-flash
    • deepseek-v4-pro
    Worth knowing
    Off-peak: All hours except peak: peak is 01:00-04:00 and 06:00-10:00 UTC, Monday to Friday, excluding Chinese public holidays. Weekends and Chinese public holidays are off-peak all day (UTC (Chinese page: peak is 09:00-12:00 and 14:00-18:00 Beijing time, weekdays)), 50% (off-peak rates are half of peak rates).
    More details
    • Search and Fetch tools: $0.75 per 1,000 successful requests each; calls made through DeepSeek Harness, Claude Code or Codex are currently free, and per-request billing starts at 00:00 Beijing time on November 1, 2026 (pricing page)
    • Old model names deepseek-v4-flash and deepseek-v4-flash-vision-exp still work and are billed at Flash prices (Chinese pricing page)
    • No subscription or coding plan was found: the pricing page lists per-token billing only. API names: deepseek-flash is DeepSeek-V4.1-Flash-1017 and deepseek-v4-pro is DeepSeek-V4-Pro-0813 (pricing page). OpenRouter's prices for DeepSeek's endpoint were read at 04:46 UTC, inside DeepSeek's off-peak hours, and match DeepSeek's off-peak rates; DeepSeek's page charges twice that at peak.
    Verified Source ↗Pricing (Chinese, CNY) ↗Get an API key
  • Inception Labs

    Inception API (Mercury models, pay-as-you-go)

    API onlyPay per tokenClear quota
    Pay per use
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    Mercury 3$0.10$0.35per 1M tokens, Cached input $0.01
    Mercury 2.5$0.04$0.15per 1M tokens, Cached input $0.004
    Mercury 2$0.25$0.75per 1M tokens, Cached input $0.025
    Mercury Decide$0.02$0per 1M tokens, Cached input $0.00

    Rates from OpenRouter, Inception's own endpoint, checked 2026-10-21

    Models
    • mercury-3
    • mercury-2.5
    • mercury-2
    • mercury-edit-2
    • mercury-decide
    More details
    • The models page lists Mercury Router with no price, API name or context length
    • The docs list per-token pricing only, and inceptionlabs.ai/pricing returned 404, so this entry makes no claim about subscriptions. Mercury Voice and Mercury Edit 2 are not listed on OpenRouter, so they have no rate here.
    Verified Source ↗Get an API key
  • Meta

    Meta Model API (pay-as-you-go)

    API onlyPay per tokenClear quota
    Pay per use
    Limits
    Metered API billing: per-token prices for each model, with no minimums or upfront commitment. Standard-mode and Contributor-mode variants have different prices. Rate limits apply per team, and each model has its own limits.
    Rules
    • Contributor variants: Meta may use prompts, code, files, context and model responses to train future models, in exchange for discounted pricing or higher usage
    • Rate limits apply per team, not per API key; for example muse-wonder-1.0 is 3,000 requests per minute and 4,000,000 tokens per minute
    API rates
    ModelInputOutputUnit and notes
    Muse Wonder$4$20per 1M tokens, Cached input $0.20
    Muse Spark 1.3$1.25$4.25per 1M tokens, Cached input $0.15
    Muse Spark 1.2$1.25$4.25per 1M tokens, Cached input $0.15
    Muse Spark 1.1$1.25$4.25per 1M tokens, Cached input $0.15
    Muse Spark 1.3 Contributor$0.10$0.20per 1M tokens, Cached input $0.002; Contributor mode: content is used to improve Meta's products
    Muse Spark 1.2 Contributor$0.10$0.20per 1M tokens, Cached input $0.002; Contributor mode: content is used to improve Meta's products

    Rates from OpenRouter, Meta's own endpoint, checked 2026-10-21

    Models
    • Muse Wonder 1.0
    • Muse Spark 1.3
    • Muse Spark 1.2
    • Muse Spark 1.1
    • Muse Spark 1.3 Contributor
    • Muse Spark 1.2 Contributor
    • Muse Image 1.0
    • Muse Voice Transcribe 1.0
    • SAM 3.1
    Works with
    • Meta Model API (direct API calls)
    More details
    • Standard mode: content is not used to train Meta models
    • Web search grounding: $2.50 per 1,000 search queries, charged in addition to token costs
    • Muse Image (muse-image-1.0): $0.01 per generated image
    • Muse Voice Transcribe (muse-voice-transcribe-1.0): $0.18 per hour of audio, metered in whole seconds
    • SAM 3.1 (sam-3.1): $2.50 per 1,000 images or $0.20 per 1,000 video frames
    • Muse Wonder Full 1.0 has no API price on the page; it is sold only by subscription in Muse Code (see the Muse Code entries). The product page lists no regions and does not say whether the API is in public preview or generally available.
    Verified Source ↗Meta Model API product page ↗Get an API key
  • MiniMax

    MiniMax API (pay-as-you-go)

    API onlyPay per tokenClear quota
    Pay per use
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    MiniMax M3$0.30$1.20per 1M tokens, Cached input $0.06

    Rates from OpenRouter, MiniMax's own endpoint, checked 2026-10-21

    Models
    More details
    • Video (pay-as-you-go, billed per second): MiniMax-H3 768P $0.08, 2K $0.13; MiniMax-H3-Max 480P $0.05, 768P $0.08; H3 regeneration 768P to 2K $0.05 per regenerated second
    • Video input images (H3): first 5 free, then $0.04 each; H3-Max: first 2 free, then $0.074 each
    • Speech (T2A, per 1M characters): speech-2.8-hd $100; speech-2.8-turbo $60; asynchronous long-form synthesis up to 1M characters per request
    • Voice design $3 per voice and rapid voice cloning $1.50 per voice, charged on first synthesis use
    • Speech-to-text (ASR) $0.38 per hour
    • Legacy per-video Hailuo models are also listed (for example MiniMax-Hailuo-2.3 768P 6s $0.28)
    • Video is not API-only: the M Plan Explore and Build subscriptions include H3 video. Video is priced per second in the API; text is priced per 1M tokens. MiniMax M3.1 Flash is listed on OpenRouter only as a free variant, so it has no paid rate here.
    Verified Source ↗Get an API key
  • Moonshot AI (Kimi)

    Kimi API (pay-as-you-go)

    API onlyPay per tokenClear quota
    Pay per use
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    Kimi K3$3$15per 1M tokens, Cached input $0.30
    Kimi K2.7 Code$0.95$4per 1M tokens, Cached input $0.19
    Kimi K2.7 Code (highspeed)$1.90$8per 1M tokens, Cached input $0.38; highspeed endpoint
    Kimi K2.6$0.95$4per 1M tokens, Cached input $0.16

    Rates from OpenRouter, Moonshot AI's own endpoint, checked 2026-10-21

    Models
    More details
    • Billing is by usage; the platform states it does not offer subscription plans and is separate from Kimi membership and Kimi Code
    • Web search is charged an additional 0.03 CNY per call that uses it (Kimi API pricing page)
    • The Kimi API platform says it sells no subscription plan; Kimi Code memberships are separate entries.
    Verified Source ↗API pricing guide ↗K3 launch post ↗Get an API key
  • Xiaomi (MiMo)

    MiMo API (pay-as-you-go)

    API onlyPay per tokenClear quota
    Pay per use
    Limits
    Pay as you go, per million tokens, at the rates below. No subscription quota.
    API rates
    ModelInputOutputUnit and notes
    MiMo-V2.6-Pro$0.435$0.87per 1M tokens, Cached input $0.0036
    MiMo-V2.6-Flash$0.14$0.28per 1M tokens, Cached input $0.0028
    MiMo-V2.6-Pro-UltraSpeed$4.35$8.70per 1M tokens, Cached input $0.036

    Rates from OpenRouter, Xiaomi's own endpoint, checked 2026-10-21

    Models
    More details
    • No time-based token discount is listed on the pay-as-you-go page; cache writes are described as limited-time free with no end date
    • TTS models (mimo-v2.5-tts and variants) are free for a limited time and are not token-priced
    • mimo-v2.5-pro and mimo-v2.5 retired at 10:00 Beijing time on October 21, 2026, per the price page. OpenRouter lists no batch rates, so batch prices are not shown.
    Verified Source ↗Get an API key