Skip to main content
API changes, such as new endpoints, parameters, errors and limits, are collected each month. Model changes are recorded from GET /models as they happen: models added, deprecated or removed, and price changes, starting October 1, 2026. Upcoming removals are also on the deprecations page, and changes to the Venice apps are in the Venice changelog. Follow this page with the RSS feed, or read it as Markdown at changelog.md.
​
ModelsDeprecations
AddedDeprecated
  • Gemma 4 Uncensored: gemma-4-uncensored, removed on November 18, 2026; use google-gemma-4-26b-a4b-it instead
Removed
  • Seedance 1.5 Pro: seedance-1-5-pro-text-to-video-basic, seedance-1-5-pro-image-to-video-basic
​
Models
Added
  • Claude Fable 5.2: claude-fable-5-2, $8.00 input and $40.00 output per 1M tokens
Removed
  • Mercury 2: mercury-2
  • Mercury 2.5: mercury-2-5
  • Mercury 3: mercury-3
​
Models
Added
  • Krea 3: krea-3, from $0.07 per image
  • Krea 3: krea-3-edit, from $0.07 per edit
​
Models
Added
​
ModelsDeprecations
AddedDeprecated
  • DeepSeek V4.1 Flash: deepseek-v4-1-flash, removed on November 15, 2026; use deepseek-v4-1-flash-1017 instead
​
Models
Added
​
DeprecationsPricing
Deprecated
  • Kimi K2.5: kimi-k2-5, removed on November 18, 2026
Price changes
  • Kimi K3.1 Pro: kimi-k3-1-pro, input $3.75 to $2.00, output $18.75 to $10.00, cache read $0.375 to $0.20 per 1M tokens
  • Kimi K3.1 Pro Highspeed: kimi-k3-1-pro-highspeed, input $7.50 to $5.00, output $37.50 to $25.00, cache read $0.75 to $0.375 per 1M tokens
​
Models
AddedRemoved
  • Grok Imagine High Quality: grok-imagine-quality-edit
  • Grok Imagine High Quality (SOTA): grok-imagine-image-quality
  • Hermes 3 Llama 3.1 405b: hermes-3-llama-3.1-405b
​
API
  • Default Model Trait — The default and function_calling_default traits now point to deepseek-v4-1-flash-1017-uncensored instead of zai-org-glm-5-2, so requests that send either trait as model, and lookups through GET /models/traits, now get the new model.
  • OpenAI Model Names — gpt-4o, gpt-4o-2024-05-13, gpt-4o-2024-08-06, gpt-4o-2024-11-20 and chatgpt-4o-latest now resolve to openai-gpt-4o-2024-11-20, and gpt-4o-mini, gpt-4o-mini-2024-07-18 and the gpt-3.5-turbo names resolve to openai-gpt-4o-mini-2024-07-18, instead of llama-3.3-70b, and are billed at the resolved model’s price. Keys restricted to private models still get llama-3.3-70b.
  • Claude Model Names — Older Claude names now resolve to Claude models instead of llama-3.3-70b or a 404: Claude 3, 3.5, 3.7 and 4.0 Sonnet names such as claude-3-5-sonnet-20241022 go to claude-sonnet-5-5, Claude 3, 3.5 and 4.5 Haiku names go to claude-haiku-5-5, and Claude 3, 4.0 and 4.1 Opus names go to claude-opus-4-8, billed at the resolved model’s price. claude-opus-4-5-20251101 is now an alias of claude-opus-4-5, and keys restricted to private models still get llama-3.3-70b for the four names it served before.
  • Image Balance Holds — /image/generate, /images/generations, /image/edit, /image/multi-edit, /image/upscale and /image/background-remove now reserve the quoted price from an API key’s balance before running, so a balance that cannot cover it gets 402 INSUFFICIENT_BALANCE up front instead of the request running and overdrawing the account. The $0.10 minimum balance no longer applies to these endpoints, and x402 wallet requests are unchanged.
  • Native Responses Mode — POST /responses now serves OpenAI models in native mode by default: instructions and text.format are applied (an invalid text.format now returns 400 instead of being ignored), output items, IDs and stream events follow the OpenAI Responses format exactly, and the Venice system prompt is off unless venice_parameters.include_venice_system_prompt is true. Send x-venice-responses-mode: compat to keep the previous behavior or native to require native mode, and read the x-venice-responses-lane response header to see which mode served a request.
  • Responses Unsupported Features — For requests to OpenAI models sent without an x-venice-responses-mode header, /responses now returns 400 for features Venice cannot serve, such as previous_response_id, conversation, stored prompts, background, the flex and priority service tiers, hosted tools like code_interpreter, file_search, image_generation and mcp, and file_id, file_url or item_reference inputs. These requests previously returned 200 with the feature silently dropped.
  • Image Content Policy Errors — /image/generate and /images/generations now return 422 when a prompt is refused for content policy, with retryable and, when available, suggested_prompt and recommended_model; some of these requests previously returned 200 with a black image. Multi-variant /image/generate requests return the variants that succeeded and report the refused ones in content_policy_violation_count and the x-venice-content-violation-count header, and content refusals on /image/edit and /image/multi-edit now return the same 422 fields.
  • Null Bytes Rejected — String fields in API requests, including chat message content and /audio/speech input, now reject the null character (U+0000) with a 400 validation error that reports String must not contain null bytes.
  • API Key IP Allowlists — POST /api_keys, PATCH /api_keys and POST /api_keys/generate_web3_key accept ipAllowlist, a list of up to 100 public IPv4 or IPv6 addresses or CIDR ranges, and key responses return it in CIDR notation. Requests from any other source IP get 403 This API key is restricted to configured IP addresses, and this request's source IP is not allowed, and an empty list allows every IP.
  • Billing Balance Credits — Earned credits are now spent on API usage after DIEM and before bundled credits and USD. GET /billing/balance adds balances.earnedCredits and balances.bundledCredits, consumptionCurrency names the first currency in that order with a positive balance (DIEM, EARNED_CREDITS, BUNDLED_CREDITS or USD), and canConsume is true when the combined spendable balance is at least $0.10. GET /billing/usage-history accepts and returns the EARNED_CREDITS currency.
  • Z-Image Turbo Rate Limits — z-image-turbo, the default image model, now allows 300 requests per minute on paid keys, up from 20.
  • Paid Tier Rate Limits — Paid keys get higher requests-per-minute limits on deepseek-v4-flash (1,000 to 2,000), deepseek-v4-1-flash (450 to 1,000), z-ai-glm-5-3-flash (1,000 to 2,000), gemma-4-uncensored (100 to 400), venice-uncensored-1-2 (150 to 300), qwen-3-8-27b (100 to 300) and jev-latest (100 to 200).
  • Transparent Image Backgrounds — /image/generate, /image/edit and /image/multi-edit accept transparent_background: true, and /images/generations now honors background: "transparent", on gpt-image-1-5, gpt-image-2-5-flare, gpt-image-2-5-sunburst and their -edit variants. GET /models reports support in model_spec.constraints.supportsTransparentBackground, and requests for other models or for JPEG output return 400.
  • Exa Web Search — POST /augment/search accepts search_provider: "exa" on every API key, with Zero Data Retention.
  • Augment Rate Limits — /augment/search and /augment/scrape now allow 60 requests per minute per user on every tier, up from 20.
  • Usage History API Key ID — Each GET /billing/usage-history entry now includes apiKeyId, the ID of the API key it was charged to, or null for web app usage and for entries that are not usage, such as purchases and grants. CSV exports add a trailing apiKeyId column.
  • Negative USD Balance — GET /billing/balance now returns a negative balances.usd when the account owes USD or a pending hold exceeds the posted USD balance, instead of null; the field is null only when the balance is zero.
  • Video Complete Changes — POST /video/complete no longer requires model, which is now deprecated and ignored, and returns 409 The media is not ready yet. Retrieve it until it has completed, then call complete. when called before the video has finished.
  • Voice Changer Access — The voice changer endpoints (/audio/voice-changer/quote, /audio/voice-changer/queue, /audio/voice-changer/retrieve and /audio/voice-changer/complete) can now be used by every API key with the elevenlabs-voice-changer model.
  • Decision Image Inputs — POST /decisions and POST /systemone accept an optional images array of 1 to 4 PNG or JPEG images, each a data: URI or a { content_type, base64 } object of up to 13,631,488 characters. Remote image URLs are not fetched.
  • Claude Tool-Loop Prompt Caching — On Claude models, earlier conversation history in tool-calling loops is now read from the prompt cache instead of being billed again at the full input rate, and cache_control set on assistant and tool messages is now honored instead of dropped; when more than four cache markers are sent, the four most recent are kept.
  • Prompt Cache Retention — /responses now accepts prompt_cache_retention (in_memory, default, extended or 24h), and on OpenAI models a value of extended or 24h on /chat/completions and /responses now requests 24-hour retention instead of being ignored. Models that do not support the field no longer return 400 when it is sent.
  • Image Detail — Image inputs on /chat/completions and /responses now pass detail (auto, low or high) to the model instead of dropping it, and accept detail: "original", which is treated as high on models that do not support it. An explicit high can increase image-token usage and cost.
  • Responses Custom Tools — In compatibility mode, /responses now applies instructions as system instructions, returns declared custom tools as custom_tool_call items with the raw input and custom-tool stream events instead of function calls, keeps the namespace of namespaced tools, and accepts a tool_choice of type custom. /responses also accepts the client_metadata field and additional_tools input items that Codex sends, which previously returned 400.
  • Responses Tool Search — /responses supports client-run tool discovery with a tool_search tool (execution: "client"), defer_loading on function tools, and tool_search_call and tool_search_output items in streamed output and replayed history. /chat/completions now returns 400 for these fields instead of ignoring them.
  • Responses Created Event — In compatibility mode, the streamed response.created event now carries the full response under response, as the OpenAI SDKs expect, and keeps its previous top-level fields, which are now deprecated.
  • Responses Beta — POST /responses is now labeled beta instead of alpha, and the API reference documents it as available to all API users.
  • Missing Model Field — API inference requests without a model field now return 400 model is required instead of 404.
  • Retry-After On Lockouts — 429 responses from the per-key error budgets for failed and unsupported-feature requests now include a Retry-After header with the whole seconds until the window resets, at least 1.
  • Error Budget Counting — Each failed response now counts once toward the failed-request budget; JSON error responses previously counted twice, which locked keys out after 25 failures in 30 seconds instead of 50.
  • Scrape Failure Status — POST /augment/scrape now returns 502 with the failure reason when a URL cannot be scraped, instead of 500.
  • Image Input Rejection Messages — When the model rejects a request’s input, /image/generate and /images/generations now return 400 with a Venice error message instead of a 500 or the model’s raw error text, and /image/edit and /image/multi-edit keep their status codes but also return Venice’s message.
  • Image Validation Order — /image/upscale and /image/background-remove now reject invalid requests with 400 before the balance and rate-limit checks, so these requests no longer use up rate-limit capacity or return 402 first. /image/background-remove now answers an image URL it cannot fetch, including private or reserved addresses, with 400 instead of 500.
  • Corrupted Image Output — When a model returns a corrupted image, the image generation and edit endpoints now return 500 The provider returned a corrupted image instead of An unknown error occurred.
  • Background Removal Size Limit — /image/background-remove now returns 422 instead of 500 for images that exceed the model’s pixel limit.
  • Image Rejection Reasons — When an image input on /chat/completions or /responses fails validation, the 400 INVALID_IMAGE_FORMAT message now ends with the specific reason, such as an image smaller than 64x64 pixels, an SVG, a missing data:image/...;base64, prefix or undecodable data.
  • Images Generations Moderation — POST /images/generations now treats moderation: null like the default auto and blurs adult images; it previously returned them unblurred, as with low.
  • Tool Message Array Content — /chat/completions tool messages whose content is an array of text parts now reach the model as plain text instead of serialized JSON, which lowers prompt tokens and stops cache_control values from leaking into the model’s input.
  • Gemini Tool Schemas — On gemini-4-argon, gemini-3-8-flash, gemini-3-7-flash, gemini-3-6-flash and gemini-3-5-flash-lite, tool parameter schemas and json_schema response formats are now passed through as written, so schemas that use $ref, $defs, oneOf, allOf, examples or type arrays such as ["string", "null"] no longer fail with 400.
  • Gemini Tool Call Streaming — On the same Gemini models, streamed tool-call arguments now arrive incrementally instead of all at once when the call completes, so long tool calls no longer time out with 504.
  • One-Token Output Caps — Requests that set max_tokens or max_completion_tokens to 1 no longer fail with 400 on kimi-k3; on some models a cap of 1 is now raised to 2 tokens, and both tokens are billed.
  • Wan 2.7 Reference Video Billing — /video/quote and charges for wan-2-7-reference-to-video now count the seconds of reference videos in addition to the output seconds; requests with only reference images are still billed on output seconds.
  • Failed Speech Requests — POST /audio/speech requests that fail with a server error are no longer charged for their input characters.
  • x402 Image Variant Charges — x402 wallet requests to /image/generate that ask for several variants are now charged for every variant returned instead of one.
  • Aliased Image Edit Billing — /image/edit and /image/multi-edit requests that name a model by an alias, such as qwen-edit, are now charged at the resolved model’s price; these edits were previously not billed.
  • Video Refunds After Delivery — POST /video/retrieve no longer refunds a video’s credits when a later retrieve call fails after the video has already been delivered.
  • Flux 3 Video Resolutions — flux-3-text-to-video, flux-3-image-to-video and flux-3-first-last-frame-to-video accept resolution values 1440p and 2160p in addition to 720p and 1080p.
  • Wan 2.7 Negative Prompts — wan-2-7-text-to-image and wan-2-7-pro-text-to-image now ignore negative_prompt.
  • MiniMax Speech Content Policy — Speech requests on tts-minimax-speech-02-hd that are refused for content policy now return 400 The request was rejected due to content policy violations. instead of 500, and are still not charged.
  • Responses Item IDs — Items that /responses returns in compatibility mode now carry IDs with the prefix for their item type, so replaying them in a later request no longer fails validation.
  • Malformed Model Suffix — Chat and Responses requests whose model feature suffix contains invalid URL encoding now return 400 instead of hanging until the client times out.
  • x402 Rate Limit Responses — x402 wallet requests to media and augment endpoints that exceed their rate limit now get 429 instead of no response.
  • API Key Revocation Errors — DELETE /api_keys now returns 500 when revocation fails, instead of returning { "success": true } while the key stays active.
  • Missing Credentials Documentation — The error reference now states that a request with no API key or x402 sign-in header returns 402 Payment Required with an x402 payment challenge, and that only an invalid API key returns 401; the API itself did not change.
  • Rate Limit Header Units — The rate limit reference now states that x-ratelimit-reset-requests and x-ratelimit-reset-tokens are Unix timestamps in milliseconds; x-ratelimit-reset-tokens was previously described as seconds until reset.
​
DeprecationsPricing
Deprecated
  • Claude Opus 4.8: claude-opus-4-8-fast, removed on November 13, 2026; use claude-opus-5-5-fast instead
Price changes
  • Kimi K3.1 Pro: kimi-k3-1-pro, input $3.00 to $3.75, output $15.00 to $18.75, cache read $0.30 to $0.375 per 1M tokens
  • Kimi K3.1 Pro Highspeed: kimi-k3-1-pro-highspeed, input $6.00 to $7.50, output $30.00 to $37.50, cache read $0.60 to $0.75 per 1M tokens
​
Models
Removed
  • Claude Sonnet 4.5: claude-sonnet-4-5
​
Models
Removed
  • Inkling: inkling
​
ModelsPricing
Removed
  • Seedance 2.5 US: seedance-2-5-us-text-to-video-private, seedance-2-5-us-image-to-video-private
  • Seedance 2.5 US R2V: seedance-2-5-us-reference-to-video-private
Price changes
​
Models
Added
​
ModelsDeprecations
Added
  • GPT-6.1 Terra: openai-gpt-61-terra, $1.00 input and $5.00 output per 1M tokens
Deprecated
  • Grok Imagine High Quality: grok-imagine-quality-edit, removed on November 2, 2026; use grok-imagine-image-2-0-edit instead
  • Grok Imagine High Quality (SOTA): grok-imagine-image-quality, removed on November 2, 2026; use grok-imagine-image-2-0 instead
  • Hermes 3 Llama 3.1 405b: hermes-3-llama-3.1-405b, removed on November 2, 2026; use deepseek-v4-1-flash instead
Removed
  • ImagineArt 1.5 Pro: imagineart-1.5-pro
​
Models
Added
  • Kling 4.0: kling-v4-text-to-video, kling-v4-image-to-video, kling-v4-reference-to-video, from $0.1963 per second
​
Models
Added
​
ModelsDeprecations
AddedDeprecated
  • Seedance 2.5 US: seedance-2-5-us-image-to-video-private, removed on October 28, 2026; use seedance-2-5-image-to-video-basic instead
  • Seedance 2.5 US: seedance-2-5-us-text-to-video-private, removed on October 28, 2026; use seedance-2-5-text-to-video-basic instead
  • Seedance 2.5 US R2V: seedance-2-5-us-reference-to-video-private, removed on October 28, 2026; use seedance-2-5-reference-to-video-basic instead
No longer scheduled for removal
​
DeprecationsModelsPricing
Deprecated
  • Inkling: inkling, removed on October 29, 2026
  • Seedance 1.5 Pro: seedance-1-5-pro-image-to-video-basic, removed on November 10, 2026; use seedance-2-0-mini-image-to-video-basic instead
  • Seedance 1.5 Pro: seedance-1-5-pro-text-to-video-basic, removed on November 10, 2026; use seedance-2-0-mini-text-to-video-basic instead
Removed
  • Qwen 3.5 9B: qwen3-5-9b
Price changes
  • Inkling: inkling, input $1.25 to $0.5625, output $5.0625 to $1.50, cache read $0.2125 to $0.125 per 1M tokens
​
ModelsDeprecations
AddedDeprecatedRemoved
  • GLM 4.7 Flash: zai-org-glm-4.7-flash
​
Models
Added
  • Flux 3 Max: flux-3-max-text-to-video, flux-3-max-image-to-video, from $0.1276 per second
  • Mercury 3: mercury-3, $0.10 input and $0.35 output per 1M tokens
  • MiniMax M3.1 Flash: minimax-m3-1-flash, $0.1875 input and $0.75 output per 1M tokens
Removed
  • Veo 3 Fast: veo3-fast-text-to-video, veo3-fast-image-to-video
  • Veo 3 Full Quality: veo3-full-text-to-video, veo3-full-image-to-video
​
Models
Added
​
Models
Added
​
ModelsPricing
Added
  • Hy4 Preview: hy4-preview, $1.0425 input and $3.12625 output per 1M tokens
  • Muse Spark Open: muse-spark-open, $1.5625 input and $5.3125 output per 1M tokens
  • Muse Wonder: muse-wonder, $5.00 input and $25.00 output per 1M tokens
Price changes
  • GPT-4o: openai-gpt-4o-2024-11-20, input $3.125 to $2.50, output $12.50 to $10.00, cache read now $1.25 per 1M tokens
  • GPT-4o Mini: openai-gpt-4o-mini-2024-07-18, input $0.1875 to $0.15, output $0.75 to $0.60, cache read $0.09375 to $0.075 per 1M tokens
  • GPT-5.2: openai-gpt-52, input $2.19 to $1.75, output $17.50 to $14.00, cache read $0.219 to $0.175 per 1M tokens
  • GPT-5.3 Codex: openai-gpt-53-codex, input $2.19 to $1.75, output $17.50 to $14.00, cache read $0.219 to $0.175 per 1M tokens
  • GPT-5.4: openai-gpt-54, input $3.13 to $2.50, output $18.80 to $15.00, cache read $0.313 to $0.25, long-context input now $5.00, long-context output now $22.50, long-context cache read now $0.50 per 1M tokens
  • GPT-5.4 Mini: openai-gpt-54-mini, input $0.9375 to $0.75, output $5.625 to $4.50, cache read $0.09375 to $0.075 per 1M tokens
  • GPT-5.4 Pro: openai-gpt-54-pro, input $37.50 to $30.00, output $225.00 to $180.00, long-context input $75.00 to $60.00, long-context output $337.50 to $270.00 per 1M tokens
  • GPT-5.5: openai-gpt-55, input $6.25 to $5.00, output $37.50 to $30.00, cache read $0.625 to $0.50, long-context input $12.50 to $10.00, long-context output $56.25 to $45.00, long-context cache read $1.25 to $1.00 per 1M tokens
  • GPT-5.5 Pro: openai-gpt-55-pro, input $37.50 to $30.00, output $225.00 to $180.00, long-context input now $60.00, long-context output now $270.00 per 1M tokens
  • GPT-5.6 Luna: openai-gpt-56-luna, input $0.25 to $0.20, output $1.50 to $1.20, cache read $0.025 to $0.02, cache write $0.3125 to $0.25, long-context input $0.50 to $0.40, long-context output $2.25 to $1.80, long-context cache read $0.05 to $0.04, long-context cache write $0.625 to $0.50 per 1M tokens
  • GPT-5.6 Luna Pro: openai-gpt-56-luna-pro, input $0.25 to $0.20, output $1.50 to $1.20, cache read $0.025 to $0.02, cache write $0.3125 to $0.25, long-context input $0.50 to $0.40, long-context output $2.25 to $1.80, long-context cache read $0.05 to $0.04, long-context cache write $0.625 to $0.50 per 1M tokens
  • GPT-5.6 Sol: openai-gpt-56-sol, input $5.00 to $4.00, output $25.00 to $20.00, cache read $0.50 to $0.40, cache write $6.25 to $5.00, long-context input $10.00 to $8.00, long-context output $37.50 to $30.00, long-context cache read $1.00 to $0.80, long-context cache write $12.50 to $10.00 per 1M tokens
  • GPT-5.6 Sol Pro: openai-gpt-56-sol-pro, input $5.00 to $4.00, output $25.00 to $20.00, cache read $0.50 to $0.40, cache write $6.25 to $5.00, long-context input $10.00 to $8.00, long-context output $37.50 to $30.00, long-context cache read $1.00 to $0.80, long-context cache write $12.50 to $10.00 per 1M tokens
  • GPT-5.6 Terra: openai-gpt-56-terra, input $2.50 to $2.00, output $15.00 to $12.00, cache read $0.25 to $0.20, cache write $3.125 to $2.50, long-context input $5.00 to $4.00, long-context output $22.50 to $18.00, long-context cache read $0.50 to $0.40, long-context cache write $6.25 to $5.00 per 1M tokens
  • GPT-5.6 Terra Pro: openai-gpt-56-terra-pro, input $2.50 to $2.00, output $15.00 to $12.00, cache read $0.25 to $0.20, cache write $3.125 to $2.50, long-context input $5.00 to $4.00, long-context output $22.50 to $18.00, long-context cache read $0.50 to $0.40, long-context cache write $6.25 to $5.00 per 1M tokens
  • GPT-6 Astra Pro: openai-gpt-6-astra-pro, input $12.50 to $10.00, output $62.50 to $50.00, cache read $1.25 to $1.00, cache write $15.625 to $12.50, long-context input $25.00 to $20.00, long-context output $93.75 to $75.00, long-context cache read $2.50 to $2.00, long-context cache write $31.25 to $25.00 per 1M tokens
  • GPT-6 Luna: openai-gpt-6-luna, input $0.125 to $0.10, output $0.625 to $0.50, cache read $0.0125 to $0.01, cache write $0.15625 to $0.125, long-context input $0.25 to $0.20, long-context output $0.9375 to $0.75, long-context cache read $0.025 to $0.02, long-context cache write $0.3125 to $0.25 per 1M tokens
  • GPT-6 Sol: openai-gpt-6-sol, input $2.50 to $2.00, output $12.50 to $10.00, cache read $0.25 to $0.20, cache write $3.125 to $2.50, long-context input $5.00 to $4.00, long-context output $18.75 to $15.00, long-context cache read $0.50 to $0.40, long-context cache write $6.25 to $5.00 per 1M tokens
  • GPT-6.1 Sol: openai-gpt-61-sol, input $2.50 to $2.00, output $12.50 to $10.00, cache read $0.125 to $0.10, cache write $3.125 to $2.50, long-context input $5.00 to $4.00, long-context output $18.75 to $15.00, long-context cache read $0.25 to $0.20, long-context cache write $6.25 to $5.00 per 1M tokens
  • GPT-6.1 Sol Ultrafast: openai-gpt-61-sol-ultrafast, input $15.00 to $12.00, output $75.00 to $60.00, cache read $0.75 to $0.60, cache write $18.75 to $15.00, long-context input $30.00 to $24.00, long-context output $112.50 to $90.00, long-context cache read $1.50 to $1.20, long-context cache write $37.50 to $30.00 per 1M tokens
​
ModelsPricing
Added
  • Gemini 4 Argon: gemini-4-argon, $2.50 input and $12.50 output per 1M tokens
Price changes
  • Claude Fable 5: claude-fable-5, input $12.00 to $10.00, output $60.00 to $50.00, cache read $1.20 to $1.00, cache write $15.00 to $12.50 per 1M tokens
  • Claude Fable 5.1: claude-fable-5-1, input $12.00 to $10.00, output $60.00 to $50.00, cache read $0.30 to $0.25, cache write $15.00 to $12.50 per 1M tokens
  • Claude Haiku 5.5: claude-haiku-5-5, input $0.125 to $0.10, output $0.625 to $0.50, cache read $0.0125 to $0.01, cache write $0.15625 to $0.125, long-context input $0.625 to $0.50, long-context output $3.125 to $2.50, long-context cache read $0.0625 to $0.05, long-context cache write $0.78125 to $0.625 per 1M tokens
  • Claude Opus 4.5: claude-opus-4-5, input $6.00 to $5.00, output $30.00 to $25.00, cache read $0.60 to $0.50, cache write $7.50 to $6.25 per 1M tokens
  • Claude Opus 4.6: claude-opus-4-6, input $6.00 to $5.00, output $30.00 to $25.00, cache read $0.60 to $0.50, cache write $7.50 to $6.25 per 1M tokens
  • Claude Opus 4.7: claude-opus-4-7, input $6.00 to $5.00, output $30.00 to $25.00, cache read $0.60 to $0.50, cache write $7.50 to $6.25 per 1M tokens
  • Claude Opus 4.8: claude-opus-4-8, input $6.00 to $5.00, output $30.00 to $25.00, cache read $0.60 to $0.50, cache write $7.50 to $6.25 per 1M tokens
  • Claude Opus 4.8: claude-opus-4-8-fast, input $12.00 to $10.00, output $60.00 to $50.00, cache read $1.20 to $1.00, cache write $15.00 to $12.50 per 1M tokens
  • Claude Opus 5: claude-opus-5, input $6.00 to $5.00, output $30.00 to $25.00, cache read $0.60 to $0.50, cache write $7.50 to $6.25 per 1M tokens
  • Claude Opus 5: claude-opus-5-fast, input $12.00 to $10.00, output $60.00 to $50.00, cache read $1.20 to $1.00, cache write $15.00 to $12.50 per 1M tokens
  • Claude Opus 5.5: claude-opus-5-5, input $4.80 to $4.00, output $24.00 to $20.00, cache read $0.24 to $0.20, cache write $6.00 to $5.00 per 1M tokens
  • Claude Opus 5.5: claude-opus-5-5-fast, input $9.60 to $8.00, output $48.00 to $40.00, cache read $0.48 to $0.40, cache write $12.00 to $10.00 per 1M tokens
  • Claude Sonnet 4.5: claude-sonnet-4-5, input $3.75 to $3.00, output $18.75 to $15.00, cache read $0.375 to $0.30, cache write $4.69 to $3.75 per 1M tokens
  • Claude Sonnet 4.6: claude-sonnet-4-6, input $3.60 to $3.00, output $18.00 to $15.00, cache read $0.36 to $0.30, cache write $4.50 to $3.75 per 1M tokens
  • Claude Sonnet 5: claude-sonnet-5, input $2.50 to $2.00, output $12.50 to $10.00, cache read $0.25 to $0.20, cache write $3.125 to $2.50 per 1M tokens
  • Claude Sonnet 5.5: claude-sonnet-5-5, input $2.50 to $2.00, output $12.50 to $10.00, cache read $0.125 to $0.10, cache write $3.125 to $2.50 per 1M tokens
​
Deprecations
Deprecated
  • GLM 4.7 Flash: zai-org-glm-4.7-flash, removed on October 20, 2026
​
Deprecations
Deprecated
  • Veo 3 Fast: veo3-fast-image-to-video, removed on October 19, 2026; use veo3.1-fast-image-to-video instead
  • Veo 3 Fast: veo3-fast-text-to-video, removed on October 19, 2026; use veo3.1-fast-text-to-video instead
  • Veo 3 Full Quality: veo3-full-image-to-video, removed on October 19, 2026; use veo3.1-full-image-to-video instead
  • Veo 3 Full Quality: veo3-full-text-to-video, removed on October 19, 2026; use veo3.1-full-text-to-video instead
​
Models
Added
​
Pricing
Price changes
  • Qwen 3.5 35B A3B: qwen3-5-35b-a3b, input $0.3125 to $0.15, output $1.25 to $1.00, cache read $0.15625 to $0.05 per 1M tokens
​
Pricing
Price changes
  • DeepSeek V4 Flash 0731: deepseek-v4-flash-0731, input $0.175 to $0.13, output $0.35 to $0.26, cache read $0.035 to $0.028 per 1M tokens
  • Flux 3: flux-3-image, 1K $0.06 to $0.12, 2K $0.145 to $0.29, 4K $0.49 to $0.98
  • Flux 3: flux-3-image-edit, per edit $0.06 to $0.12, 1K $0.06 to $0.12, 2K $0.145 to $0.29, 4K $0.49 to $0.98
​
ModelsDeprecationsPricing
AddedDeprecated
  • DeepSeek V3.2: deepseek-v3.2, removed on October 21, 2026
  • Qwen 3.5 9B: qwen3-5-9b, removed on October 21, 2026
Removed
  • Background Remover: bria-bg-remover
Price changes
​
Models
Added
​
Pricing
Price changes
  • DeepSeek V4.1 Flash: deepseek-v4-1-flash, input $0.375 to $0.30, output $1.50 to $1.20 per 1M tokens
​
ModelsPricing
Added
  • Flux 3: flux-3-image, from $0.06 per image
  • Flux 3: flux-3-image-edit, from $0.06 per edit
Price changes
  • Claude Sonnet 5: claude-sonnet-5, input $3.00 to $2.50, output $15.00 to $12.50, cache read $0.30 to $0.25, cache write $3.75 to $3.125 per 1M tokens
  • Claude Sonnet 5.5: claude-sonnet-5-5, input $3.75 to $2.50, output $18.75 to $12.50, cache read $0.375 to $0.25, cache write $4.6875 to $3.125 per 1M tokens
​
ModelsDeprecationsPricing
AddedDeprecated
  • Claude Sonnet 4.5: claude-sonnet-4-5, removed on October 30, 2026
Price changes
  • Mercury 2.5: mercury-2-5, input $0.05 to $0.25, output $0.1875 to $0.9375, cache read $0.005 to $0.025 per 1M tokens