# ByteSpike

> Anthropic-compatible AI gateway with one key for 36 frontier models. Public per-model rate card. Failures don't bill.

## Docs

- [Admin operations runbook · ByteSpike admin](https://docs.bytespike.ai/admin-operations.md): A one-page operations guide for admins: grant employees model permissions, monitor usage, configure models, billing, emergency-response SOPs — every action has its exit point marked.
- [GET /me/account](https://docs.bytespike.ai/api-reference/account/me-account.md): Your account dashboard in one call — balance, allowed models, quotas, profile.
- [GET /me/billing/transactions](https://docs.bytespike.ai/api-reference/account/me-billing-transactions.md): Wallet ledger — every top-up, refund, and balance adjustment, paginated.
- [GET /me/notifications (+ unread-count / mark-read / mark-all-read)](https://docs.bytespike.ai/api-reference/account/me-notifications.md): In-app notification feed. Polls the same data the console bell shows.
- [GET /me/usage](https://docs.bytespike.ai/api-reference/account/me-usage.md): Per-day / per-model spend rollup. Powers the console usage chart.
- [POST /me/webhooks (+ list / update / delete / deliveries)](https://docs.bytespike.ai/api-reference/account/me-webhooks.md): Subscribe an HTTPS endpoint to your account events. Five endpoints in one round-trip family.
- [GPT-4o Image](https://docs.bytespike.ai/api-reference/image/gpt-4o-image.md): GPT-4o's image-output mode — vision model that generates images directly inside the chat completions response.
- [GPT-Image 2](https://docs.bytespike.ai/api-reference/image/gpt-image-2.md): OpenAI's photorealistic image generator — strong on prompts that mix description with explicit composition rules.
- [GPT-Image 2 Official](https://docs.bytespike.ai/api-reference/image/gpt-image-2-official.md): The compliance-tier variant of GPT-Image 2 for audit- and data-residency-sensitive workflows.
- [Nano Banana](https://docs.bytespike.ai/api-reference/image/nano-banana.md): Google's photorealistic image generator — studio-grade lighting and material rendering, the entry point to the Nano Banana family.
- [Nano Banana Pro](https://docs.bytespike.ai/api-reference/image/nano-banana-pro.md): Nano Banana with extended detail pass and material refinement — the photoreal flagship for hero imagery.
- [Nano Banana 2](https://docs.bytespike.ai/api-reference/image/nano-banana-v2.md): Latest Nano Banana iteration — sharper in-image text rendering and improved hand / face anatomy.
- [Seedream V4](https://docs.bytespike.ai/api-reference/image/seedream-v4.md): ByteDance flagship aesthetic image generator — 1024² text-to-image with native CJK prompt understanding.
- [Seedream V4.5](https://docs.bytespike.ai/api-reference/image/seedream-v4-5.md): ByteDance's 2048² aesthetic image generator — refined edges around text and packaging, native CJK prompt understanding.
- [Seedream V5 lite](https://docs.bytespike.ai/api-reference/image/seedream-v5lite.md): Lightweight Seedream variant — cheap iteration at 1024², the right pick for prompt design and rough exploration.
- [API Reference](https://docs.bytespike.ai/api-reference/overview.md): Every endpoint, one base URL, one key.
- [Claude Haiku 4.5](https://docs.bytespike.ai/api-reference/text/claude-haiku-4-5.md): Anthropic's smallest, fastest 4-series model. The default for production agents that have to chain dozens of LLM calls without choking on latency.
- [POST /messages](https://docs.bytespike.ai/api-reference/text/claude-messages.md): Anthropic-native Messages endpoint — Claude family + cross-vendor models behind the same shape.
- [Claude Opus 4.5](https://docs.bytespike.ai/api-reference/text/claude-opus-4-5.md): First 4-series Opus — 200K context at Opus quality. Strong on long-form reasoning, superseded by 4.6 / 4.7 / 4.8 for new work.
- [Claude Opus 4.6](https://docs.bytespike.ai/api-reference/text/claude-opus-4-6.md): Refined 4-series Opus — 200K context with tighter long-form output and reduced hallucination on multi-document synthesis.
- [Claude Opus 4.7](https://docs.bytespike.ai/api-reference/text/claude-opus-4-7.md): Prior-generation Opus, superseded by Claude Opus 4.8 — 200K context, top-of-line reasoning, the model to reach for when one shot has to be right.
- [Claude Opus 4.8](https://docs.bytespike.ai/api-reference/text/claude-opus-4-8.md): The current flagship Opus — 200K context, top-of-line reasoning. The model to reach for when one shot has to be right.
- [Claude Sonnet 4.5](https://docs.bytespike.ai/api-reference/text/claude-sonnet-4-5.md): The mid-tier Sonnet that displaced 3.5 — strong on tool use and structured output without the Opus latency.
- [Claude Sonnet 4.6](https://docs.bytespike.ai/api-reference/text/claude-sonnet-4-6.md): The current production Sonnet — same context as 4.5 with measurably better tool use and code generation. Default Anthropic mid-tier.
- [DeepSeek V3.2](https://docs.bytespike.ai/api-reference/text/deepseek-v3-2.md): DeepSeek's last 3-series flagship — strong on code generation, kept for benchmark-validated paths.
- [DeepSeek V4 Flash](https://docs.bytespike.ai/api-reference/text/deepseek-v4-flash.md): Latency-optimised V4 — DeepSeek Pro quality on bounded prompts at sub-second TTFB.
- [DeepSeek V4 Pro](https://docs.bytespike.ai/api-reference/text/deepseek-v4-pro.md): DeepSeek's flagship V4 — strongest code generation in the open-weight tier, with reasoning chain support.
- [Gemini 3.1 Pro](https://docs.bytespike.ai/api-reference/text/gemini-3-1-pro.md): Google's flagship Gemini — 1M context, native multimodal, deep reasoning. The model to reach for when input is long AND mixed-modality.
- [Gemini 3 Flash](https://docs.bytespike.ai/api-reference/text/gemini-3-flash.md): Google's small / fast Gemini 3 — high-throughput vision-native classification and extraction.
- [POST /v1beta/models/{model}:generateContent](https://docs.bytespike.ai/api-reference/text/gemini-generate-content.md): Gemini-native protocol — drop-in for the Google Generative AI SDKs.
- [GLM-5](https://docs.bytespike.ai/api-reference/text/glm-5.md): Zhipu's GLM-5 — open-weight Chinese flagship with strong tool use and structured output.
- [GLM-5.1](https://docs.bytespike.ai/api-reference/text/glm-5-1.md): Zhipu's current flagship — refined GLM-5 with vision support and tighter tool-call generation.
- [GPT-5.2](https://docs.bytespike.ai/api-reference/text/gpt-5-2.md): Mid-tier 5-series — the standard model from the first 5.x refinement wave, kept for benchmark-validated paths.
- [GPT-5.4](https://docs.bytespike.ai/api-reference/text/gpt-5-4.md): 5.4 standard tier — the workhorse 5-series for general production work, balancing latency and quality.
- [GPT-5.4 mini](https://docs.bytespike.ai/api-reference/text/gpt-5-4-mini.md): 5.4-tier mini — the small model with 5.4's tighter tool-call generation. Drop-in cost-down from GPT-5.4.
- [GPT-5.4 nano](https://docs.bytespike.ai/api-reference/text/gpt-5-4-nano.md): 5.4-tier nano — sub-second routing with 5.4's structured output gains. The fastest 5.4-series model.
- [GPT-5.4 pro](https://docs.bytespike.ai/api-reference/text/gpt-5-4-pro.md): GPT-5.4 with reasoning_effort dial — the right call for multi-step problems where the first draft has to compose, not just answer.
- [GPT-5.5](https://docs.bytespike.ai/api-reference/text/gpt-5-5.md): OpenAI's current production flagship — 128K context, native reasoning, the model to pick for any new project that doesn't have a benchmark for an older version.
- [GPT-5.5 instant](https://docs.bytespike.ai/api-reference/text/gpt-5-5-instant.md): Latency-tuned variant of GPT-5.5 — same flagship quality on shorter prompts, no reasoning chain delay.
- [GPT-5 mini](https://docs.bytespike.ai/api-reference/text/gpt-5-mini.md): OpenAI's small 5-series — the right default for production extraction, structured output, and lightweight agent steps.
- [GPT-5 nano](https://docs.bytespike.ai/api-reference/text/gpt-5-nano.md): OpenAI's smallest 5-series model — sub-second latency for routing, classification, and high-throughput agent loops.
- [Kimi K2.5](https://docs.bytespike.ai/api-reference/text/kimi-k2-5.md): Moonshot's K2.5 — long-context Chinese-language flagship, strong on extraction and document QA.
- [Kimi K2.6](https://docs.bytespike.ai/api-reference/text/kimi-k2-6.md): Current Moonshot flagship — 128K context with measurably tighter output and tool-call generation.
- [MiniMax M2.5](https://docs.bytespike.ai/api-reference/text/minimax-m2-5.md): MiniMax's prior flagship — strong on creative writing and dialogue, kept for benchmark-validated paths.
- [MiniMax M2.7](https://docs.bytespike.ai/api-reference/text/minimax-m2-7.md): MiniMax's current flagship — refined M2.5 with longer context, vision support, and stronger general-purpose reasoning.
- [POST /chat/completions](https://docs.bytespike.ai/api-reference/text/openai-chat-completions.md): OpenAI Chat Completions shim — every model behind a single OpenAI-shape request.
- [POST /responses](https://docs.bytespike.ai/api-reference/text/openai-responses.md): OpenAI Responses API on ByteSpike — get o-series and GPT-5 structured-output shape against any model in the catalog.
- [Reading your balance](https://docs.bytespike.ai/api-reference/utility/balance.md): How to check remaining credits, rate-limit budgets, and per-call spend — there's no separate /v1/balance endpoint; the data lives on every response and in /api/v1/me/*.
- [POST /tasks/cancel](https://docs.bytespike.ai/api-reference/utility/tasks-cancel.md): Abort a non-terminal async task. Idempotent — calling on an already-finished task returns the current state.
- [POST /tasks/query](https://docs.bytespike.ai/api-reference/utility/tasks-query.md): Read the current state of an async task by task_id or out_task_id. Free.
- [POST /tasks/submit](https://docs.bytespike.ai/api-reference/utility/tasks-submit.md): Kick off an async multimodal task (image, video, …) and get a task_id back immediately. The wait happens in the background.
- [GET /models](https://docs.bytespike.ai/api-reference/utility/v1-models.md): List the models your key can call, with capability tags and live pricing pointers.
- [Seedance 1.5 Pro](https://docs.bytespike.ai/api-reference/video/seedance-1-5-pro.md): ByteDance's prior video flagship — strong on character motion and Chinese-market briefs, kept for benchmark-validated paths.
- [Seedance 2.0](https://docs.bytespike.ai/api-reference/video/seedance-2-0.md): ByteDance Seedance 2.0 standard tier — the new mid-tier video generator with improved character consistency and CJK prompt handling.
- [Seedance 2.0 Pro](https://docs.bytespike.ai/api-reference/video/seedance-2-0-pro.md): Seedance 2.0 with extended detail pass — ByteDance's video flagship for hero-quality character motion.
- [Seedance 2.0 Pro Fast](https://docs.bytespike.ai/api-reference/video/seedance-2-0-pro-fast.md): Seedance 2.0 Pro fidelity at half the wait — the latency-tuned Pro variant for time-bound hero work.
- [Seedance Fast](https://docs.bytespike.ai/api-reference/video/seedance-fast.md): Speed floor of the Seedance family — sub-30s turnaround on 5s clips for prompt iteration and time-bound UX.
- [Sora 2](https://docs.bytespike.ai/api-reference/video/sora-2.md): OpenAI's text-to-video flagship — 1080p, async via tasks API.
- [Sora 2 Pro](https://docs.bytespike.ai/api-reference/video/sora-2-pro.md): Sora 2 at higher fidelity, longer max duration, and full 1080p — the model when the clip is the deliverable.
- [Veo 3.1](https://docs.bytespike.ai/api-reference/video/veo-3-1.md): Google's text-to-video flagship — strong on natural-world footage and physics-coherent motion.
- [Veo 3.1 Fast](https://docs.bytespike.ai/api-reference/video/veo-3-1-fast.md): Veo 3.1 tuned for ~30s turnaround — lower fidelity, faster iteration on natural-world footage.
- [Authentication](https://docs.bytespike.ai/authentication.md): API key format, the three protocol headers, key management, and the per-key controls (group binding, quota, rate limits, IP allowlist).
- [Claude Code CLI](https://docs.bytespike.ai/claude-code-cli.md): Point Anthropic's Claude Code CLI at ByteSpike. Two env vars, done — works against every Anthropic-compatible model in the catalog.
- [Codex CLI](https://docs.bytespike.ai/codex-cli.md): Point OpenAI's Codex CLI at ByteSpike. Works against the GPT-5 family, the o-series, and every Claude / Gemini model reachable via Responses-API translation.
- [Credits & Billing](https://docs.bytespike.ai/concepts/credits-and-billing.md): How credits work, what bills (and doesn't), how to read the accounting headers, and where to manage quotas.
- [DOSIA Agent mode](https://docs.bytespike.ai/concepts/dosia-agent-mode.md): How DOSIA Agent uses ByteSpike — which protocol it speaks, which models it can target, what cache_control and tool_use blocks it expects to pass through, and the per-model caveats DOSIA users should know.
- [Endpoint types](https://docs.bytespike.ai/concepts/endpoints.md): ByteSpike exposes five real endpoint shapes behind one key — three text protocols (Anthropic Messages, OpenAI Chat Completions, OpenAI Responses, Gemini Native) plus two multimodal shapes (Image, Async Tasks). Pick the protocol your client already speaks.
- [Error handling](https://docs.bytespike.ai/concepts/error-handling.md): The retry / failure mental model — when to retry, when not, and the no-charge guarantee on every non-2xx.
- [Errors reference](https://docs.bytespike.ai/concepts/errors-reference.md): Every HTTP status, error code, and envelope shape you'll see from ByteSpike. Per-status retry policy. No charges on any non-2xx.
- [Legacy compat names](https://docs.bytespike.ai/concepts/legacy-compat-names.md): Old DeepSeek model strings (deepseek-chat, deepseek-reasoner) still resolve through ByteSpike until DeepSeek's official 2026/07/24 sunset. New code should target the canonical names — deepseek-v4-flash and deepseek-v4-pro — directly.
- [MCP integration · DOSIA bridge](https://docs.bytespike.ai/concepts/mcp-integration.md): How DOSIA talks to ByteSpike over MCP — one OAuth connect, then your main brain can generate images, write with external LLMs, and produce video without any of that wiring being your problem.
- [Multimodal](https://docs.bytespike.ai/concepts/multimodal.md): How image, video, audio, and embedding endpoints sit alongside text under the same key.
- [Rate limits & concurrency](https://docs.bytespike.ai/concepts/rate-limits.md): Per-key rate-limit buckets (5h / 1d / 7d), concurrency, headers to read, and what 429 looks like when you hit a wall.
- [Streaming](https://docs.bytespike.ai/concepts/streaming.md): Server-Sent Events across the three protocols. How to enable, parse, abort, and handle mid-stream failures.
- [Configure your client](https://docs.bytespike.ai/configure-client.md): After signing up — wire any client to ByteSpike: base URL, API key, protocol choice, model name.
- [Cursor IDE](https://docs.bytespike.ai/cursor.md): Add ByteSpike as an OpenAI-compatible provider in Cursor — works with the editor's chat, agent, and tab-complete features.
- [DOSIA](https://docs.bytespike.ai/dosia.md): DOSIA is ByteSpike's first-party desktop AI client. Two modes: personal (you bring a ByteSpike key) or enterprise (admin-provisioned, lark SSO, auto-allowed models).
- [Employee onboarding · DOSIA × ByteSpike](https://docs.bytespike.ai/employee-onboarding.md): Get every team member through the DOSIA → ByteSpike loop in 5 minutes: install, switch to enterprise mode, enterprise SSO, auto-pick up allowed models, start chatting / vision / image generation.
- [Gemini CLI](https://docs.bytespike.ai/gemini-cli.md): Point Google's official Gemini CLI at ByteSpike. Works against gemini-3-1-pro / gemini-3-5-flash / gemini-2-5-flash plus every model reachable via Gemini Native translation.
- [Introduction](https://docs.bytespike.ai/introduction.md): One key, one base URL, every model — Anthropic-compatible by default.
- [Migrate from Anthropic](https://docs.bytespike.ai/migrate-from-anthropic.md): Switch from Anthropic's API to ByteSpike with two env vars. Same Anthropic SDK, same Messages shape, every Claude + cross-vendor model in our catalog.
- [Migrate from OpenAI](https://docs.bytespike.ai/migrate-from-openai.md): Switch from OpenAI's API to ByteSpike with three changes: base URL, API key, model id. Same OpenAI SDK, same code, every model in our catalog.
- [claude-haiku-4-5](https://docs.bytespike.ai/models/claude-haiku-4-5.md): Anthropic's small-and-cheap model — for classification, routing, structured extraction at scale. Vision and tool use included; pair with prompt caching on the system prompt for the lowest possible per-request cost.
- [claude-opus-4-7](https://docs.bytespike.ai/models/claude-opus-4-7.md): Anthropic's prior-generation Opus, superseded by claude-opus-4-8. 200K context, deep reasoning, long-document review, multi-document synthesis. Extended thinking, web search, and vision all supported.
- [claude-opus-4-8](https://docs.bytespike.ai/models/claude-opus-4-8.md): Anthropic's flagship Claude — codebase-scale reasoning, multi-document synthesis, and hard agentic work with a 200K context window. Extended thinking, web search, and vision all supported. Released 2026-05-29.
- [claude-sonnet-4-6](https://docs.bytespike.ai/models/claude-sonnet-4-6.md): Anthropic's production workhorse mid-tier. Vision, tool use, prompt caching, extended thinking, and web search — together at Sonnet pricing. Available through the Anthropic Messages protocol and OpenAI Chat Completions.
- [deepseek-r1](https://docs.bytespike.ai/models/deepseek-r1.md): DeepSeek R1 reasoning model. Explicit chain-of-thought output, structured for downstream consumers. Best paired with `reasoning_effort` to control depth vs latency.
- [deepseek-v3-2](https://docs.bytespike.ai/models/deepseek-v3-2.md): DeepSeek V3.2 — the older V3 series of the DeepSeek family. Chat-Completions-only, 64K context. For the latest V4 family with reasoning chain, see deepseek-v4-pro.
- [deepseek-v3-anthropic](https://docs.bytespike.ai/models/deepseek-v3-anthropic.md): DeepSeek V3.2 served through the Anthropic Messages protocol surface. Same DeepSeek-V3-2 model, but you keep your Anthropic SDK, your tool_use blocks, and your cache_control markers.
- [deepseek-v4-flash](https://docs.bytespike.ai/models/deepseek-v4-flash.md): DeepSeek V4 Flash — small-mid tier of DeepSeek's V4 family. Anthropic and OpenAI-compatible. The best Chinese-LLM dollar-per-quality at the moment, well-suited to high-volume classification and agent loops.
- [deepseek-v4-pro](https://docs.bytespike.ai/models/deepseek-v4-pro.md): DeepSeek V4 Pro — the cost-effective reasoning flagship. A fraction of the cost of gpt-5-4 at comparable quality on most benchmarks. Dual endpoint (Chat Completions + Anthropic Messages). Reasoning chain exposed. Vision not on HTTP API.
- [doubao-seed-2.0-pro](https://docs.bytespike.ai/models/doubao-seed-2-0-pro.md): ByteDance's flagship Doubao Seed 2.0 Pro — vision, tools, long context. Strong on Chinese-language tasks, available through OpenAI Chat Completions.
- [gemini-2-5-flash](https://docs.bytespike.ai/models/gemini-2-5-flash.md): Google's Gemini 2.5 Flash — small-mid multimodal with 1M context window. Surprisingly capable for general chat at Haiku-comparable pricing.
- [gemini-3-1-pro](https://docs.bytespike.ai/models/gemini-3-1-pro.md): Google's Gemini 3.1 Pro — current flagship Gemini. Native vision-on-image and vision-on-video input, 1M context, broad tool support. The default Google-stack vision model and DOSIA `analyze_video` companion.
- [gemini-3-5-flash](https://docs.bytespike.ai/models/gemini-3-5-flash.md): Google's Gemini 3.5 Flash — fast multimodal with a 1M-token context window. The step up from 3-flash for harder reasoning while staying well below Pro pricing.
- [gemini-3-flash](https://docs.bytespike.ai/models/gemini-3-flash.md): Google's Gemini 3 Flash — fast, low-cost multimodal with a 200K context window. Great for high-volume chat, vision, and grounded fresh-fact tasks.
- [glm-5.1](https://docs.bytespike.ai/models/glm-5.md): Zhipu's GLM-5.1 — the published Chinese-LLM flagship for general chat with vision support. Available through OpenAI Chat Completions.
- [gpt-5-2](https://docs.bytespike.ai/models/gpt-5-2.md): OpenAI's mid-tier GPT-5 with built-in web search and tool-use chaining. Step above 5-4-mini, step below 5-4 — the right pick when you need browsing or longer tool chains but not the flagship reasoning budget.
- [gpt-5-4](https://docs.bytespike.ai/models/gpt-5-4.md): OpenAI's GPT-5.4 — the production workhorse of the GPT-5 family. Vision, tools, reasoning, web search, and structured outputs together at mid-tier pricing. Available through both Chat Completions and the Responses API.
- [gpt-5-4-mini](https://docs.bytespike.ai/models/gpt-5-4-mini.md): OpenAI's small-mid GPT — production-grade reasoning at a fraction of gpt-5-4's price. Use it for classification, routing, structured extraction, and short-prompt agent loops.
- [gpt-5-4-nano](https://docs.bytespike.ai/models/gpt-5-4-nano.md): OpenAI's smallest GPT-5.4 tier — the cheapest GPT-flavored reasoning on the gateway. Use it for very-high-volume classification, routing, and short structured-extraction calls.
- [gpt-5-5](https://docs.bytespike.ai/models/gpt-5-5.md): OpenAI's GPT-5.5 — current flagship of the GPT-5 family. Vision-capable, reasoning-tunable via `reasoning_effort`. The right default for non-Anthropic main brain work or for DOSIA `chat_with` external-LLM calls.
- [gpt-image-2](https://docs.bytespike.ai/models/gpt-image-2.md): OpenAI's flagship image model. Text-to-image and image-to-image (mask-driven edits) on one endpoint, billed per image. The default choice when you want OpenAI fidelity on layout, typography, and product shots.
- [Models](https://docs.bytespike.ai/models/index.md): Every model available on ByteSpike, organized by family. Each model page lists its public price and protocol, and a runnable quickstart.
- [kimi-k2-6](https://docs.bytespike.ai/models/kimi-k2-6.md): Moonshot's Kimi K2.6 — long-context Chinese-LLM with a 128K-token context window. Available through OpenAI Chat Completions.
- [minimax-m2-7](https://docs.bytespike.ai/models/minimax-m2-7.md): MiniMax M2.7 — competitive Chinese-LLM general chat. Available through OpenAI Chat Completions.
- [nano-banana](https://docs.bytespike.ai/models/nano-banana.md): Google's lightweight image model. Text-to-image at materially lower per-image cost than gpt-image-2, with the same OpenAI Images protocol. The volume / cost-sensitive default.
- [nano-banana-v2](https://docs.bytespike.ai/models/nano-banana-v2.md): Google's lightweight image model, second generation. Adds image-to-image edits while keeping the nano-banana cost profile. The default volume-tier model for any pipeline that needs to mutate an input photo.
- [seedance-1.5-pro](https://docs.bytespike.ai/models/seedance-1-5-pro.md): ByteDance Seedance 1.5 Pro — the previous-generation Seedance flagship, kept on the rate card for compatibility. New work should target [`seedance-pro`](/models/seedance-pro).
- [seedance-pro](https://docs.bytespike.ai/models/seedance-pro.md): ByteDance Seedance Pro — current Seedance flagship video model. Strong on Chinese aesthetic and East-Asian subject quality. The default Seedance tier for new projects.
- [seedance-pro-fast](https://docs.bytespike.ai/models/seedance-pro-fast.md): Seedance Pro on a faster runtime — same flagship model, lower wall-clock to first frame, lower per-second price. The choice when you're iterating Seedance prompts and don't want to wait.
- [seedance2](https://docs.bytespike.ai/models/seedance2.md): ByteDance Seedance 2 — the newer Seedance generation video model. Strong Chinese aesthetic and East-Asian subject handling, with a faster low-cost variant in seedance2-fast.
- [seedance2-fast](https://docs.bytespike.ai/models/seedance2-fast.md): Seedance 2 Fast — the lightest Seedance 2 tier. Lower fidelity than seedance2, materially lower per-second cost, fastest wall-clock in the Seedance family. The default volume-tier choice on the ByteDance video stack.
- [seedream-4](https://docs.bytespike.ai/models/seedream-4.md): ByteDance's Seedream 4 flagship image model. Strong on Chinese aesthetic, poster composition, and East-Asian product photography. Lower per-image cost than OpenAI; richer in-image typography support than nano-banana.
- [seedream-4.5](https://docs.bytespike.ai/models/seedream-4-5.md): ByteDance Seedream's refined 4.5 release — tighter prompt adherence, cleaner edge fidelity, slightly higher per-image cost than seedream-4. The default upgrade target if seedream-4 was already serving your use case.
- [seedream-v5lite](https://docs.bytespike.ai/models/seedream-v5lite.md): ByteDance Seedream's lightweight tier. The lowest per-image cost in the image lineup, traded against tighter generation quality on hard prompts. The default for high-volume programmatic image work where good-enough beats expensive.
- [sora2](https://docs.bytespike.ai/models/sora2.md): OpenAI's Sora 2 video model. Text-to-video and image-to-video on a task-based polling endpoint. Billed per second of generated footage. The default choice when you want Sora-grade physics, camera motion, and shot composition.
- [sora2-pro](https://docs.bytespike.ai/models/sora2-pro.md): OpenAI Sora 2 Pro — longer clips (up to 30s), 1080p, richer motion dynamics and subject persistence over [`sora2`](/models/sora2). Highest per-second cost in the video lineup; the choice when clip duration and consistency matter more than budget.
- [veo3.1](https://docs.bytespike.ai/models/veo3-1.md): Google's Veo 3.1 video model. Native audio generation alongside the video track is the key differentiator from sora2 — useful when your clip is the final output rather than a layer that gets sound design pass.
- [veo3.1-fast](https://docs.bytespike.ai/models/veo3-1-fast.md): Google Veo 3.1 Fast — the lower-cost tier of Veo. Trades some fidelity and the audio option for a meaningfully cheaper per-second price. Default for volume / programmatic video pipelines on the Google stack.
- [Pricing](https://docs.bytespike.ai/pricing.md): Per-model rate card across the 40+ model catalog — chat, image, video. Refreshed nightly.
- [Quickstart](https://docs.bytespike.ai/quickstart.md): Make your first request in under two minutes.
- [Register an account](https://docs.bytespike.ai/register.md): Create a ByteSpike account in ~30 seconds. Email + password, or Google sign-in. Trial credits arrive on signup — no card required.
- [Top up credits](https://docs.bytespike.ai/top-up.md): Buy ByteSpike credits via Stripe — minimum $5, up to +11.2% bonus on bigger packs.
- [Troubleshooting · ByteSpike × DOSIA](https://docs.bytespike.ai/troubleshooting.md): One-stop troubleshooting for employees / admins — install / SSO / model permissions / balance / abuse / emergency-response SOP all in one place.