Skip to main content
gemini-2-5-flash is Google’s small-mid Gemini — fast, multimodal, with the 1M-token context window that distinguishes the entire Gemini family. Competitive with gpt-5-4-mini and claude-haiku-4-5 on most benchmarks, plus the long-context advantage. Pricing: 0.50/1Minput,0.50 / 1M input, 3.00 / 1M output, $0.05 / 1M cache read — see the rate card.

​
Protocols

ProtocolPath
Gemini NativePOST https://llm.bytespike.ai/v1beta/models/gemini-2-5-flash:generateContent
OpenAI Chat Completions (shim)POST https://llm.bytespike.ai/v1/chat/completions
Anthropic Messages (translated)POST https://llm.bytespike.ai/v1/messages
For the OpenAI shim, the gateway translates the request body to Gemini’s generateContent shape behind the scenes. From the client side, you write standard openai-SDK code.

​
Quickstart

curl https://llm.bytespike.ai/v1/chat/completions \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "gemini-2-5-flash",
    "messages": [{ "role": "user", "content": "Hello, ByteSpike." }]
  }'

​
Capabilities

CapabilitySupported
Chat Completions (shim)✅
Streaming (SSE)✅
Vision (image input)✅
Tools / function calling✅ parallel
JSON mode✅
Grounding (web search)✅
Long context✅ 1M tokens
Context window1M tokens

​
When to use

  • Long-context cheap — 1M context at Haiku price is unique. Codebase reviews, multi-doc QA where 200K isn’t enough.
  • Fresh-fact tasks at low cost — grounding (Google Search) is supported on Flash.
  • High-volume vision — OCR-style tasks on receipts / invoices / screenshots.
When not to use:
  • Hard reasoning — gemini-3-1-pro is sized for it; Flash is more chat-tier.
  • Anthropic Messages protocol — Gemini only speaks the OpenAI shim today.

​
Next