Claude Opus 4.8

Vendor: Anthropic Model ID: claude-opus-4-8 Capability: 200K context · tool use · vision · prompt caching · streaming · extended thinking Pricing: per-token, Opus tier (live rate) Opus 4.8 is the current flagship Anthropic model and the successor to Opus 4.7. It is the model you reach for when the one shot has to be right. It’s slower than Sonnet, more expensive than Sonnet, and noticeably better at the things Sonnet starts cutting corners on: long-context reasoning, multi-step plans where each step depends on the last, and the kind of code generation where the first draft has to compile and match the architecture conventions of an existing codebase. With extended thinking enabled, the response wait grows but the answer quality on hard problems jumps further than the latency cost suggests.

Request

curl https://llm.bytespike.ai/v1/messages \
  -H "x-api-key: $BYTESPIKE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "max_tokens": 16384,
    "messages": [
      {"role": "user", "content": "Implement an LRU cache with O(1) get and put."}
    ]
  }'

Body parameters

Field	Type	Required	Default	Notes
`model`	string	yes	—	`claude-opus-4-8`
`messages`	array	yes	—	Conversation history. Up to 200K tokens of input.
`max_tokens`	integer	yes	—	Hard cap. Max for this model: 32768.
`system`	string \| array	no	—	Array form supports `cache_control`.
`temperature`	number	no	1.0	Range 0.0–1.0.
`top_p`	number	no	1.0	Nucleus sampling.
`tools`	array	no	—	Supported, parallel calls supported.
`tool_choice`	object	no	`{"type":"auto"}`	`auto` / `any` / `tool` (named).
`thinking`	object	no	—	Extended-thinking. Higher budget = better long-reasoning answer at higher latency.
`stream`	boolean	no	false	SSE streaming.

Response

{
  "id": "msg_opus_…",
  "type": "message",
  "role": "assistant",
  "model": "claude-opus-4-8",
  "content": [
    {"type": "thinking", "thinking": "<extended reasoning trace>"},
    {"type": "text", "text": "Here's the LRU cache..."}
  ],
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 32,
    "output_tokens": 1248,
    "thinking_tokens": 4032
  }
}

thinking_tokens are billed at the input-token rate (extended thinking adds latency but not the full output cost). See the pricing table for current rate.

Code examples

curl https://llm.bytespike.ai/v1/messages \
  -H "x-api-key: $BYTESPIKE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "max_tokens": 16384,
    "messages": [{"role": "user", "content": "Implement an LRU cache with O(1) get and put."}]
  }'

import anthropic

client = anthropic.Anthropic(
    base_url="https://llm.bytespike.ai/v1",
    api_key="$BYTESPIKE_API_KEY",
)

message = client.messages.create(
    model="claude-opus-4-8",
    max_tokens=16384,
    messages=[
        {"role": "user", "content": "Implement an LRU cache with O(1) get and put."}
    ],
)

# When extended thinking is on, content[0] is the thinking block; the
# text block is the last one in the array.
text_block = next(b for b in message.content if b.type == "text")
print(text_block.text)

import Anthropic from "@anthropic-ai/sdk"

const client = new Anthropic({
  baseURL: "https://llm.bytespike.ai/v1",
  apiKey: process.env.BYTESPIKE_API_KEY,
})

const message = await client.messages.create({
  model: "claude-opus-4-8",
  max_tokens: 16384,
  messages: [
    {
      role: "user",
      content: "Implement an LRU cache with O(1) get and put.",
    },
  ],
})

const textBlock = message.content.find((b) => b.type === "text")
console.log(textBlock.text)

Extended thinking

Opt in by setting the thinking block:

{
  "model": "claude-opus-4-8",
  "max_tokens": 16384,
  "thinking": {
    "type": "enabled",
    "budget_tokens": 8192
  },
  "messages": [...]
}

budget_tokens is the maximum number of internal-reasoning tokens. The model may use fewer; the floor is a few hundred. Recommended budgets:

Task	Suggested budget
Multi-step coding	4K–8K
Long-context summarisation	8K–16K
Hard math / proof	16K–32K

Higher budgets monotonically improve answer quality on hard problems — but the marginal return falls off above 16K for most tasks.

Cache control

{
  "model": "claude-opus-4-8",
  "system": [
    {
      "type": "text",
      "text": "<the corpus you keep referring to>",
      "cache_control": {"type": "ephemeral"}
    }
  ],
  "messages": [...]
}

Cache reads at the discounted rate visible in the pricing table. On Opus 4.8, cache control is the single highest-leverage cost optimisation — large system prompts paid once, billed at the cache-read rate on every subsequent turn.

Errors

Code	Trigger	Billed?
400	Body validation failed	No
401	Missing / revoked key	No
402	Wallet exhausted (Opus calls trip this faster)	No
413	Input exceeds 200K tokens	No
429	Rate-limited	No
5xx	Upstream provider issue	No (auto-retry envelope)

When to use

One-shot quality matters and you can wait for a thoughtful answer.
Code generation in an existing codebase where conventions matter.
Multi-step plans where each step depends on the last (Sonnet starts skipping; Opus 4.8 keeps the chain tight).
Long-context reasoning across legal / medical / technical corpora within the 200K window.
For mid-tier cost / latency, see Sonnet 4.6.
For high-throughput agent loops, see Haiku 4.5.

Limits

Limit	Value
Context window	200K tokens
Max output	32768 tokens
Supports tool use	Yes (parallel)
Supports vision	Yes
Supports streaming	Yes
Supports prompt caching	Yes
Supports extended thinking	Yes

API Reference

Protocol endpoints

OpenAI

Anthropic

DeepSeek

Moonshot

Zhipu

MiniMax

Image

Video

Account & billing

Utility

Request

Body parameters

Response

Code examples

Extended thinking

Cache control

Errors

When to use

Limits

​Request

​Body parameters

​Response

​Code examples

​Extended thinking

​Cache control

​Errors

​When to use

​Limits

Request

Body parameters

Response

Code examples

Extended thinking

Cache control

Errors

When to use

Limits