GPT-5 mini - ByteSpike

Vendor: OpenAI Model ID: gpt-5-mini Capability: 128K context · tool use · vision · streaming · structured output Pricing: per-token, mini tier (live rate) GPT-5-mini is the small model that displaced GPT-4o-mini for new production work. Same price tier, measurably tighter structured output, better tool-call argument generation. For most extraction and routing flows, this is the starting point — only step up to a standard 5-series model when you’ve benchmark-confirmed that mini’s quality plateau is the bottleneck.

Request

curl https://llm.bytespike.ai/v1/chat/completions \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "gpt-5-mini",
    "messages": [
      {"role": "user", "content": "Extract the dates from: The deal closes 2024-08-12 with backup date 2024-09-01."}
    ],
    "response_format": {"type": "json_object"}
  }'

Body parameters

Field	Type	Required	Default	Notes
`model`	string	yes	—	`gpt-5-mini`
`messages`	array	yes	—	OpenAI chat shape.
`max_tokens`	integer	no	model max	Max: 16384.
`temperature`	number	no	1.0	Range 0.0–2.0.
`tools`	array	no	—	Function calling supported (parallel).
`tool_choice`	string \| object	no	`"auto"`	—
`response_format`	object	no	—	JSON mode + structured output (recommended for extraction).
`stream`	boolean	no	false	SSE streaming.

Response

{
  "id": "chatcmpl-…",
  "object": "chat.completion",
  "model": "gpt-5-mini",
  "choices": [{"index": 0, "message": {"role": "assistant", "content": "{\"dates\": [\"2024-08-12\", \"2024-09-01\"]}"}, "finish_reason": "stop"}],
  "usage": {"prompt_tokens": 32, "completion_tokens": 18, "total_tokens": 50}
}

Code examples

curl https://llm.bytespike.ai/v1/chat/completions \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "gpt-5-mini",
    "messages": [{"role": "user", "content": "Extract the dates as JSON."}],
    "response_format": {"type": "json_object"}
  }'

Streaming + caching

"stream": true for SSE. Automatic prompt caching — keep system prompt and tool schema stable for max cache hit rate.

Errors

Code	Trigger	Billed?
400	Body validation	No
401	Missing / revoked key	No
402	Wallet exhausted	No
422	Param not supported	No
429	Rate-limited	No
5xx	Upstream provider issue	No

When to use

Production extraction / structured output / routing.
Lightweight agent steps (one tool call per step).
For higher quality, see GPT-5.5 or GPT-5.4.
For lower latency, see GPT-5-nano.
For 5.4-era mini, see GPT-5.4-mini.

Limits

Limit	Value
Context window	128K tokens
Max output	16384 tokens
Supports tool use	Yes (parallel)
Supports vision	Yes
Supports streaming	Yes
Supports prompt caching	Automatic
Supports structured output	Yes

​Request

​Body parameters

​Response

​Code examples

​Streaming + caching

​Errors

​When to use

​Limits