HopBase DocsModel changelog
Model launches, retirements, and list-price changes, newest first.
Current rates are in the signed-in model catalog; the full list of IDs on sale is in Available models.
Retirement policy
Retirements are announced on this page. When the model maker gives advance notice of a retirement, we pass it on here as soon as we receive it. After retirement the ID is removed from GET /v1/models, requests for it return 404 model_not_found, and the replacement is named in the entry below. Change the model string to the replacement; the Base URL, key, and request format stay the same unless the entry says otherwise.
Changes
| Date | Model IDs | Change | Replacement / notes |
|---|---|---|---|
| 2026-11-07 | MiniMax-H3-Max | Capability update: reference-to-video. Up to 9 reference images, 3 reference videos (2–15 s each, ≤ 15 s in total, .mp4 / .mov) and 3 reference audio tracks (only alongside an image or video reference), the same limits as MiniMax-H3; first/last frames and references are mutually exclusive. Still 480P / 768P and 5–15 s, no 2K. Output price unchanged ($0.05 / s at 480P, $0.08 / s at 768P); reference video is billed separately per input second ($0.0553 / s at 480P, $0.143 / s at 768P); the first 2 input images per task are free, then $0.074 per image; audio is free. MiniMax-H3-Max is now also served by the MiniMax H3 Official group, alongside MiniMax-H3 | MiniMax Hailuo |
| 2026-11-07 | kling-lip-sync | Request change: each extra.face_choose item now requires sound_start_time, sound_end_time and sound_insert_time (non-negative integer milliseconds, end > start), which crop the audio and place it on the face's timeline within the face's detected StartTime–EndTime; sound_volume and original_audio_volume (0–2) are optional | Kling lip sync |
| 2026-11-06 | gpt-image-2.6-flare, gpt-image-2.6-sunburst | Added (GPT Image 2.6, released by OpenAI on 2026-11-04; Flare is speed-focused, Sunburst quality-focused). Same official price as 2.5; low / medium emit more output tokens than on 2.5. New: the complexity parameter, SVG output, animated GIF output, and markup-guided edits (markup) | GPT Image |
| 2026-10-30 | deepseek-v4.1-flash-1017-sg | Added (DeepSeek V4.1 Flash, 2026-10-17 release, served from a Singapore data center in its own group, DeepSeek Singapore Data Center, which any user can add to a key). Same official price and peak / off-peak schedule as deepseek-v4.1-flash-1017: weekdays 09:00–12:00 and 14:00–18:00 Beijing time are peak, every other hour is half price. 1M context, context caching, tool calling and thinking; works over Chat Completions, Responses and Anthropic Messages | DeepSeek Singapore |
| 2026-10-19 | deepseek-v4.1-flash-1017 | Added (DeepSeek V4.1 Flash, 2026-10-17 release). Same official price and peak / off-peak schedule as deepseek-v4.1-flash: ¥2 / ¥0.04 / ¥8 (input / cache hit / output per 1M tokens) at peak, half off peak. 1M context, up to 393,216 output tokens, reads images, calls tools, and works over Chat Completions, Responses, and Anthropic Messages. deepseek-v4.1-flash stays on the 0910 release and remains available; both versions run side by side | DeepSeek V4.1 / V4 |
| 2026-10-15 | kimi-k3 | Capability update: the thinking trace is now available. Send "include_reasoning": true to get it in reasoning_content (streaming: delta.reasoning_content, before the answer); without it only the answer is returned. Thinking tokens bill at the output rate and count toward max_tokens | Kimi K3 |
| 2026-10-15 | seed-icl-2.0 | Added (Doubao Voice Cloning 2.0, in the same group as Doubao Speech). Send a recording to POST /v1/voice_clone to clone a voice (returns a voice_id starting with hbv_), then synthesize on POST /v1/audio/speech with model: seed-icl-2.0 and that voice_id; list and delete with POST /v1/get_voice / POST /v1/delete_voice. Synthesis is billed at the official price of ¥3 per 10,000 characters (same as seed-tts-2.0), and the optional preview text is billed per character at the same price; cloning, listing, and deleting are free, there is currently no cloning fee, and failed requests are not billed. During the beta each account can keep one cloned voice | Doubao voice cloning |
| 2026-10-14 | doubao-seed-2-1-pro-260915, doubao-seed-2-1-lite-260915 | Added (Doubao Seed chat, same group as the existing five). Official price (input / output / cache hit per 1M tokens, CNY): 2.1 Pro (260915) ¥6 / ¥30 / ¥1.2; 2.1 Lite ¥0.8 / ¥2.7 / ¥0.16; neither is tiered by input length. Both have a 1M context and up to 256K output, read images, call tools, think by default, and work over Chat Completions, Responses, and Anthropic Messages. The existing doubao-seed-2-1-pro-260628 (256K context) stays available | Available models |
| 2026-10-14 | seed-tts-2.0, seed-audio-1.0 | Added (Doubao Speech, both in one group). seed-tts-2.0 is text to speech at an official ¥3 per 10,000 characters; seed-audio-1.0 generates audio with tone and sound effects from a scene description at an official ¥1 per minute of generated audio. OpenAI-compatible POST /v1/audio/speech only, input up to 5,000 / 3,000 characters, no streaming; failed requests are not billed | Text-to-speech API |
| 2026-10-14 | doubao-seed-2-1-pro-260628, doubao-seed-2-1-turbo-260628, doubao-seed-evolving, doubao-seed-character-260628, doubao-seed-2-0-mini-260428 | Added (Doubao Seed chat, all five in one group). Official price (input / output / cache hit per 1M tokens, CNY): 2.1 Pro and Evolving ¥6 / ¥30 / ¥1.2; 2.1 Turbo ¥3 / ¥15 / ¥0.6; Character ¥0.8 / ¥2 / ¥0.16 for input up to 32K and ¥1.2 / ¥6 / ¥0.16 for 32K–128K; 2.0 Mini ¥0.2 / ¥2 / ¥0.04, ¥0.4 / ¥4 / ¥0.08 and ¥0.8 / ¥8 / ¥0.16 for input up to 32K, 32K–128K and 128K–256K. All read images and call tools over Chat Completions, Responses, and Anthropic Messages | Available models |
| 2026-10-14 | gemini-4-argon | Added (Gemini 4 Argon). Official introductory price through 2026-11-30: $2 / $10 (input / output per 1M tokens; output includes thinking), cache read $0.10; from 2026-12-01: $4 / $20, cache read $0.20. 1M context, no long-context tier | Gemini chat |
| 2026-10-14 | All Gemini chat models on the Gemini Official Direct (Enterprise) group | Capability update: /v1/responses and /v1/messages added (Claude Code connects directly through CC Switch); tool calling, structured output, stop, and n work on all three protocols; input_audio and file parts are accepted, and unsupported fields and parts now return 400 | Gemini chat |
| 2026-10-12 | MiniMax-H3-SelfHosted | Added: self-hosted Hailuo H3 (H3-level quality at a lower price). 480P / 768P / 1080P / 2K / 4K, 4-15 s; reference video and audio must be public URLs | MiniMax Hailuo H3 video API |
| 2026-10-10 | gemini-nano-banana-2.1 | Added (Nano Banana 2.1). 1K / 2K / 4K: $0.0336 / $0.0504 / $0.1134 per image, about half the price of gemini-3.1-flash-image; the ID has no -image suffix | Gemini image generation |
| 2026-10-08 | claude-haiku-5-5 | Added. Tiered by prompt length: $0.10 / $0.50 (input / output per 1M tokens) up to 100K prompt tokens, $0.50 / $2.50 above 100K | Anthropic SDK |
| 2026-09-29 | claude-sonnet-5-5 | Added | Anthropic SDK |
| 2026-09-29 | claude-sonnet-5 | Price change: corrected down to the current official rate, $2 / $10 (input / output per 1M tokens) | Anthropic SDK |
| 2026-09-23 | claude-opus-5-5 | Added | Anthropic SDK |
| 2026-09-23 | gpt-6-sol, gpt-6-luna | Added | OpenAI SDK |
| 2026-09-23 | speech-2.8-hd, speech-2.8-turbo | Added: voice cloning | Text-to-speech API |
| 2026-09-21 | midjourney-v8-2 | Added | Midjourney image API |
| 2026-09-20 | claude-sonnet-4-5-20250929 | Retired | claude-sonnet-5 |
| 2026-09-20 | claude-opus-4-5-20251101 | Retired | claude-opus-5-5 |
| 2026-09-17 | glm-5.3-flash | Added | GLM-5.3 Flash |
| 2026-09-17 | deepseek-v4.1-flash | Price change: billed at the official peak / off-peak rates | DeepSeek V4.1 / V4 |
| 2026-09-17 | MiniMax-H3, MiniMax-H3-Max | Price change: aligned with the official international list price | MiniMax Hailuo H3 video API |
| 2026-09-16 | speech-2.8-hd, speech-2.8-turbo | Added | Text-to-speech API |
| 2026-09-11 | grok-imagine-image-2.0 | Price change: generation price lowered to the official rate | Grok Imagine images and video |
| 2026-09-10 | deepseek-v4.1-flash | Added | DeepSeek V4.1 / V4 |
| 2026-09-09 | gpt-image-2.5-flare, gpt-image-2.5-sunburst | Added | Image API |
| 2026-09-06 | doubao-seedance-2-0-fast-260128-a, doubao-seedance-2-0-mini-260615-a, doubao-seedance-2-5-260628-a | Added | Video generation |
| 2026-09-05 | gpt-6-astra | Added | OpenAI SDK |
| 2026-09-03 | gemini-3.8-flash | Added | Gemini chat |
| 2026-09-03 | gemini-3.6-flash | Price change: corrected down to the official rate | Gemini chat |
| 2026-09-02 | claude-fable-5-1 | Added | Anthropic SDK |
| 2026-08-31 | MiniMax-H3-Max | Added | MiniMax Hailuo H3 video API |
| 2026-08-28 | gemini-3.7-flash | Added | Gemini chat |
| 2026-08-26 | qwen3.8-max, qwen3.8-flash, qwen3.7-max, qwen3.7-plus, qwen3.7-flash | Added | Qwen |
| 2026-08-25 | deepseek-v4-pro-202606 | Added | DeepSeek V4.1 / V4 |
| 2026-08-25 | deepseek-v4-flash-202605 | Price change: raised to follow the official increase of 2026-08-16; billed at the peak rate at all hours | DeepSeek V4.1 / V4 |
| 2026-08-25 | deepseek-v4-flash (ID without date) | Retired | deepseek-v4-flash-202605 |
| 2026-08-24 | wan3.0-video, happyhorse-1.1-t2v, happyhorse-1.1-i2v, happyhorse-1.1-r2v | Added | Wan 3.0 and HappyHorse video API |
| 2026-08-24 | grok-imagine-video-1.5 | Added | Grok Imagine images and video |
| 2026-08-19 | kimi-k3 | Added | Kimi K3 |
| 2026-08-19 | glm-5.2 | Retired | glm-5.3 — see GLM-5.3 |
| 2026-08-14 | glm-5.3 | Added | GLM-5.3 |
| 2026-08-14 | MiniMax-H3 | Added | MiniMax Hailuo H3 video API |
Changes before 2026-08-14 are not listed.
Next steps
- Available models: find the model IDs available to a key
- Error codes and retries: what each error code means and whether to retry
- Account and catalog API: check the balance, usage, and model catalog over the API