Deep Dive
AI Radar
Telegram
Global users
Media Models
New additions
🚀 Claude Opus 4.7 is live on API易! Anthropic’s new flagship: +13% over Opus 4.6 on a 93-task coding benchmark, 3× production tasks on Rakuten-SWE-Bench, tool errors cut to 1/3, new xhigh effort tier.
claude-opus-4-7-thinking variant also supported, pricing stays at $5/$25 per 1M tokens. View Details📊 We’ve created a dedicated Nano Banana pricing comparison page — covering default pricing, standard deposit (10% bonus), and VIP deposit (20% bonus) tiers, so you can maximize savings on top of base prices. Stability comes first — 24/7 high-concurrency, production-grade quality, so you can focus on building your product. Nano Banana Pricing Guide
💰 Due to rising official API key costs and high-speed CDN bandwidth for image downloads, we’ve carefully considered a modest adjustment to Nano Banana 2 token-based pricing, effective April 20 (UTC+8): input $0.14→$0.18/M tokens, output $16.80→$21.6/M tokens, official discount raised to 36%. Modest increase on a low base, stackable with deposit bonus up to 20% — great for daily usage. Deposit Promotions
✅ Nano Banana Pro and Nano Banana 2 ran smoothly all day today. The default channel is fully restored with consistent call logs. Feel free to use in production — for higher volume, you can scale up directly.
✅ Default channel has restored 
gemini-3-pro-image-preview (Nano Banana Pro) supply, fully operational! Price adjusted to $0.09/call, NanoBananaEnterprise rate lowered to 1.4x. Best practice: Default channel ($0.09/call) as primary, NanoBananaEnterprise (1.4x, stable HA) as fallback. We’re committed to long-term stable supply — speed and high availability are our edge. Deposit $100+ for 10% bonus, more deposits = bigger bonus! Deposit Promotions🍌 Good news! Nano Banana Pro now has a dedicated low-cost channel
NanoBananaCheap at $0.09/image per-call, under $0.08 with deposit bonus. Edit your existing token to set NanoBananaCheap as default — takes effect instantly without disrupting live services. Note: Default channel Nano Banana Pro pricing will adjust to $0.09/call from April 14, 0:00 (UTC+8) to ensure long-term stable Gemini image model supply.🍌 Nano Banana 2 now has more resources, default channel working normally. Nano Banana Pro still only available via Enterprise channel (
NanoBananaEnterprise) with per-call billing. Coming soon: lower-priced model names gemini-3-pro-image-preview-2k and gemini-3-pro-image-preview-1k, stay tuned.✅ Model status:
gemini-3.1-pro-preview text multimodal series running normally; gemini-2.5-flash-image (Mini Banana) normal and on special today 🍌🍌 Default channel still under investigation, not yet restored. For urgent use, enable the
NanoBananaEnterprise channel. Tutorial🍌 Nano Banana 2: Default channel temporarily restored but still congested. Nano Banana Pro: currently only available via Enterprise HA channel (2.5x) with per-call billing, ~¥0.79/image (for users with $100+ deposit bonus).
🍌 Nano Banana token setup best practice: set billing mode to “Token Priority” (compatible with Nano Banana 2 token-based + Nano Banana Pro per-call), default channel 
Default (1x), fallback channel NanoBananaEnterprise (2.5x, ~50% off official). Traffic only falls back to Enterprise when Default fails — ensuring zero downtime. Acceptable for personal use; for B2C products with fixed pricing, please evaluate cost impact.🍌 Default channel Nano Banana series is down (IP proxy issue + KEY quota exhausted), fix in progress. Temporary workaround: edit your token to set
NanoBananaEnterprise as fallback channel, or switch directly to it (2.5x rate).🚀
glm-5.1 Zhipu’s most powerful open-source coding Agent is live! SWE-Bench Pro 58.4 beats GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro. 744B MoE, 200K context, autonomous 8-hour long-horizon tasks. 📖 Learn more🎬 Veo 3.1 video generation timing out: traced to slowdowns on
flow.google itself — our tests didn’t return videos even after 10 minutes. This is a Google-side issue, please try again later.✅ After 30 minutes of monitoring, the ultra-low-price default channels for Nano Banana Pro / 2 are running normally since 10:00 (UTC+8). Safe to continue using.
🍌 Our Google Nano Banana series uses direct official API. Keys dropped faster than usual this morning — please retry failed tasks. You can edit your token and select the
NanoBananaEnterprise channel as fallback at 2.5x pricing (still ~50% off official), stable with no 429s.🍌 New
NanoBananaEnterprise channel is live — The Most Stable API for Nano Banana Series! 4K Banana Pro at 50% off official pricing, ample concurrency, no 429 errors. Recommended setup: default channel as primary, this channel as fallback for rock-solid reliability.🚀
qwen3.6-plus Alibaba’s Qwen3.6 series first model is live! MoE 72B params, Terminal-Bench 61.6 surpassing Claude Opus 4.5, 1M context, China’s top coding Agent. 📖 Learn more🍌 New model alias
gemini-3.1-flash-image-preview-4k — Nano Banana 2 with 4K output, ready for Cherry Studio, Chatbox, LobeChat and other AI clients. Use “per-request billing” token for a flat $0.055/image.💰 Claude cache write pricing reduced from 1.25x to 1x! Since multi-account rotation may cause 5min cache misses, we lowered the write multiplier to balance costs for users.
🍌 Both Nano Banana Pro (
gemini-3-pro-image-preview) and Nano Banana 2 (gemini-3.1-flash-image-preview) are running stable — full steam ahead!💡 Fallback tip:
gemini-3-pro-preview text model is resource-constrained. Set a reasonable timeout, then fall back to gemini-3-flash-preview to ensure business continuity.🍌 Stability update:
gemini-3.1-flash-image-preview (Nano Banana 2) running normally; gemini-3-pro-image-preview (Nano Banana Pro) under resource pressure — retry on 429. Set timeout to 360s, no need to wait longer.✅ Nano Banana Pro and Nano Banana 2 ran stable all day — a good start to April! 🍌
🚀
gemini-embedding-2-preview Google’s first native multimodal embedding model is live! Unified vectors for text/image/video/audio/PDF, MTEB English #1 at 68.32, text just $0.20/M tokens. 📖 Learn more🍌 Nano Banana Pro is also affected by Google’s 503 high-demand error — concurrency is limited, awaiting Google’s recovery. We also experienced a brief 502 caused by task backlog from Google’s slowdown leading to memory overflow; the container has auto-restarted and service is back to normal.
🍌 Nano Banana 2 (
gemini-3.1-flash-image-preview) is returning Google’s error: “This model is currently experiencing high demand.” This is a Google-side compute issue, not a platform problem. We recommend setting timeout to 360 seconds to handle slow responses during high load.🚀
seed-2-0-pro-260328 ByteDance Seed 2.0 Pro flagship reasoning model is now live! AIME 2025 at 98.3, Codeforces 3020, ~4-6x cheaper than GPT-5.2. 📖 Learn more🍌 Gemini has occasional instability recently. If
gemini-3-pro-image-preview times out or fails, 👉 use gemini-3.1-flash-image-preview as a fallback. Nano Banana Pro is $0.09/call fixed price; Nano Banana 2 is even cheaper.✅ Nano Banana Pro and Nano Banana 2 are both back to normal speed! Recovered around 4:00 AM (UTC+8). Welcome to continue using them 🍌
🍌 Nano Banana Pro: Google still hasn’t fixed it — waiting for recovery. Nano Banana 2 (
gemini-3.1-flash-image-preview): completely unavailable. We’ve paused routing to reduce server load — you’ll see “no available channel” which is expected. We’ll re-enable it as soon as it’s back.🍌 OG Nano Banana gemini-2.5-flash-image is working normally;
gemini-3.1-pro-preview text multimodal model is also fine — our account status is healthy. Nano Banana Pro and Nano Banana 2 compute appears only ~50% recovered, still monitoring.⚠️
gemini-3.1-flash-image-preview and gemini-3-pro-image-preview are returning Google’s official error: “This model is currently experiencing high demand.” This is a Google-side compute overload — we’re monitoring closely.🍌
gemini-3-pro-image-preview (Nano Banana Pro) appears to have recovered and is back to normal. Welcome to continue using it!🍌 Nano Banana 2 is currently more stable than Nano Banana Pro. Pro is affected by Google’s official compute capacity issues — response times are slow, awaiting Google’s recovery. Monitor status via
openrouter.ai/google/gemini-3-pro-image-preview/uptime (green line = official compute).🍌
gemini-3.1-flash-image-preview has recovered after 1 PM (UTC+8) and is running normally. Welcome to send traffic — keep using Nano Banana 2! 🍌🎬 Sora 2 budget reverse-proxy version has very low success rate — suspected official deprecation in progress. Last successful call at 13:00 (UTC+8). Sora 2 official-relay group and Veo 3.1 are working normally.
📖 What Base URLs are available on APIYI? We offer 4 nodes: Global Direct, China Default, Backup/Business, and Cloudflare CDN. See details 👉 Base URL Configuration Guide
🍌 Nano Banana 2 (
gemini-3.1-flash-image-preview) is experiencing limited concurrency — some requests may get stuck occasionally. We’re actively working on a fix. Thanks for your patience.📖 New FAQ added: Base URL Configuration Guide — explains the differences between OpenAI
/v1, Claude root domain, and Gemini /v1beta endpoints, with code examples and error troubleshooting.📖 New FAQ added: Model Multiplier & Pricing Guide — explains how multipliers work and billing logic for different model types. We’re continuously improving our docs to provide clear answers for every common question.
🚀 MiniMax-M2.7 and MiniMax-M2.7-highspeed are now available! M2.7: $0.3/1M input tokens, $1.2/1M output tokens. M2.7-highspeed: $0.6/1M input tokens, $2.4/1M output tokens. Pay-per-use Chat billing.
✅ The frequent 429 rate-limiting issue on the dashboard has been resolved. Page refresh, logs, and top-up functions are back to normal.
🎬 Sora-2 budget version is running stable today — give it a try! Use the default or sora-2 group, $0.12 per generation.
⚠️ Dashboard may show 429 rate-limiting due to high overall traffic — this is a self-protection mechanism and resolves within minutes. API calls are NOT affected. We’re working on a fix.
🍌 Both Nano Banana Pro (
gemini-3-pro-image-preview) and Nano Banana 2 (gemini-3.1-flash-image-preview) are running stable today. Happy weekend!⚠️ Nano Banana 2 (
gemini-3.1-flash-image-preview) is under some load, slower than usual. Nano Banana Pro (gemini-3-pro-image-preview) is working great — recommended as the primary choice.🚀 4 new Grok 4.20 Beta models now available:
grok-4.20-beta, grok-4.20-multi-agent-beta-0309, grok-4.20-beta-0309-non-reasoning, grok-4.20-beta-0309-reasoning. Official pricing ($2 input / $6 output per M tokens), discounts via deposit promotions. Learn more✅ Veo 3.1 series video generation API is back to normal — restored around 3:00 PM (UTC+8) today and running stable.
✅ Google capacity is back — Nano Banana Pro (
gemini-3-pro-image-preview) has been rock-solid since 6:30 PM (UTC+8) tonight, generating 2K–4K images in 30s to 1 min.
⚠️ Nano Banana Pro (
gemini-3-pro-image-preview) is under high load again due to Google capacity shifts. We recommend also integrating gemini-3.1-flash-image-preview for flexible switching or as a user-selectable option.✅ Nano Banana Pro (
gemini-3-pro-image-preview) faced very high traffic this afternoon — we’ve added more servers and keys, and it’s now stabilized. Continuing to monitor.💡 Tip: Due to Google’s capacity fluctuations, we recommend setting
timeout to 300 seconds for Nano Banana Pro and 2 APIs. An occasional retry may be needed — a clean timeout is better than hanging indefinitely without output.⚠️ Nano Banana Pro (
gemini-3-pro-image-preview) is again hit by Google’s high demand issue: This model is currently experiencing high demand. This is on Google’s side. Nano Banana 2 remains unaffected.🍌🍌 Nano Banana Pro (
gemini-3-pro-image-preview) and Nano Banana 2 (gemini-3.1-flash-image-preview) are both running rock-solid with plenty of concurrency — fire away!⚠️ Veo 3.1 low-cost reverse-proxy version is temporarily unavailable — Google’s
labs.google/fx/tools/flow is down, causing video tasks to appear successful but fail to deliver videos. Affected credits will be refunded tomorrow; please contact support if any are missed.✅ Nano Banana Pro (
gemini-3-pro-image-preview) has been restored to normal levels — still under active monitoring. Google has tightened resource controls recently, making keys prone to suspension. Most competitors are down; our recovery was hard-won.⚠️ Nano Banana Pro (
gemini-3-pro-image-preview) is experiencing capacity issues — we’re adding more keys to restore concurrency ASAP. Nano Banana 2 (gemini-3.1-flash-image-preview) is working normally.🚀 Now available:
gpt-5.4-mini and gpt-5.4-nano — lightweight, affordable models priced the same as OpenAI official, with ~10% off via top-up promotions. Great for OpenClaw and similar use cases.🎬 Sora-2 low-cost reverse-proxy version has a low success rate but can still produce results — account pool optimization is ongoing. The official Sora-2 relay version remains fully operational for those who need stability.
🎬 Sora-2 low-cost reverse-proxy version is back online — still monitoring, hopefully not another one-day wonder. Use it while you can! Usage guide
📊 Nano Banana Pro (
gemini-3-pro-image-preview) and Nano Banana 2 are mostly normal this afternoon despite very high traffic — we’re holding up well. Occasional errors from Google’s side: This model is currently experiencing high demand. This is a Google capacity issue, not a full outage like yesterday.🚀 Recommended:
gemini-3.1-flash-lite-preview is now available! Fast and affordable — ideal for translation, summarization, and other high-frequency tasks.✅ Nano Banana Pro is mostly back to normal — about 3% of requests may still experience brief delays as Google restores capacity. Overall availability is good.
Nano Banana Pro (
gemini-3-pro-image-preview) is gradually recovering but still intermittently returning: This model is currently experiencing high demand. Nano Banana 2 (gemini-3.1-flash-image-preview) is working normally now.📢 Google image model status update:
gemini-3.1-flash-image-preview is slow but still functional; gemini-3-pro-image-preview is severely down on Google’s end. By comparison, text models and 1st-gen Nano Banana (gemini-2.5-flash-image) are running normally, but most users rely on v2 and Pro. We are monitoring the situation.⚠️ Nano Banana 2 is returning 503 errors due to high demand on Google’s side. Original error:
This model is currently experiencing high demand. Spikes in demand are usually temporary. Please try again later. Currently slower but still functional, pending Google’s capacity recovery. Pro models are also affected, with generation times roughly doubled (~100s for 4K images).Weekend special! To compensate for the ~30-minute outage on March 13, default and SVIP user groups enjoy a 0.95x discount this Saturday and Sunday (UTC+8) — now live! Groups already at 0.88x (e.g. DeepSeek, Qwen) are excluded. Have a great weekend!
✅ OVH outage resolved — API Yi website and API calls are fully restored. This incident affected all services globally. We will publish a post-incident report and a compensation policy soon. Please retry any failed tasks from the past 30 minutes. Thank you for your patience! 🙏
⚠️ Urgent: Our server provider OVH is experiencing a system outage. The issue is being resolved and service is expected to recover within 15 minutes. Thank you for your patience.
The website dashboard is experiencing 429 rate-limiting issues. API calls are not affected, but dashboard access may be impacted. Temporary workaround: please use the backup domain https://b.apiyi.com/token instead.
The OpenClaw documentation has been refined to help you get started faster — let’s raise some “shrimp” together! Learn more
Google has deprecated gemini-3-pro-preview (shut down March 9). We’ve synced accordingly — calls to this model now redirect to gemini-3.1-pro-preview automatically. No impact on your services. Note: this is a text multimodal model, not the Nano Banana image model series.
We’ve compiled common reasons for image generation failures with the Nano Banana Image API series. Check it out if you run into issues. Learn more
The frequent 429 errors have been resolved. During the 1-2 minute container restart window, 502 errors may occur — these failed requests are not billed, but tasks will fail and need to be resubmitted. We apologize for the inconvenience.
The dashboard is experiencing frequent 429 errors. API calls are not affected, but the dashboard and top-up functions may be impacted (for manual top-up, contact us on Enterprise WeChat). We are working on a fix — try switching IP addresses as a temporary workaround.
seed-2-0-lite-260228 is now available! ByteDance’s latest multimodal text model from Volcano Engine / BytePlus. Learn more
All Gemini models have returned to normal levels. gemini-3.1-flash-image-preview has been allocated more resources and is now more stable.
Veo 3.1 reverse proxy: Google suddenly added a faint watermark (bottom-right corner) to generated videos today. Investigating a fix — no ETA yet.
GPT-5.4 and GPT-5.4-Pro are now available. Note: Pro version is expensive — use with caution. Not available for free-tier users.
gemini-3-pro-image-preview running normally; gemini-3.1-flash-image-preview overloaded. New models have unstable capacity early on — prefer Pro over Flash.
Sora 2 budget version short-lived — blocked by risk control again, currently unavailable. Sora 2 official API works normally. View official API docs
Sora 2 reverse proxy restored — $0.12 per generation, an affordable official video solution. Live status here
gemini-3-pro-image-preview and gemini-3.1-flash-image-preview resources replenished. Both running smoothly with fast response times.
Veo 3.1 video model may return “internal service error” due to Google Flow tightening captcha verification, causing occasional task failures — please retry. Supply remains sufficient; Veo 3.1 is a solid alternative while Sora 2 is unavailable.
gemini-3-pro-image-preview had concurrency issues around 10:00 AM, resolved after 10:20 AM. Recommend using gemini-3-pro-image-preview over gemini-3.1-flash-image-preview, as the latter has high load from the provider. Official API keys face frequent risk control — maintaining current stability is no easy feat.
gemini-3-pro-image-preview normal, gemini-3.1-flash-image-preview high load from provider