Anthropic: Claude Opus 4.8 (batch)

Claude Opus 4.8 (batch) is a text generation model from Anthropic, offered through api14 under the model id anthropic/claude-opus-4.8:batch. It sits in the Opus family, which Anthropic describes as its most capable generally available line, and this particular variant is the batch mode of that model. If you are already working within Anthropic's model naming conventions, the Opus tier is the one positioned above the smaller Claude tiers in terms of capability.

The model accepts text, image, and file inputs and produces text output. It has a context window of 1,000,000 tokens, and the maximum output length is 128,000 tokens. Reasoning is supported, and the provider description confirms this multimodal input setup alongside text-only output. In practice this means you can send large documents, long conversation histories, or mixed content that includes images and files, and receive generated text that reasons over all of it within a single request.

Those capabilities point to a fairly specific set of use cases. The million-token context suits work where you need to keep an entire codebase, a long contract, a research corpus, or a multi-hour transcript in view at once rather than chunking it. The file and image inputs make it useful for document analysis, screenshot or diagram understanding, and workflows that mix prose with visual or attached material. Because output is text only, it fits generation, summarisation, extraction, question answering, and agent-style tool use rather than image or audio production. Reasoning support helps on problems that benefit from step-by-step deliberation before an answer.

To call it through api14, use the OpenAI-compatible chat completions interface with the model id anthropic/claude-opus-4.8:batch. The request parameters available to you are include_reasoning, max_tokens, reasoning, reasoning_effort, response_format, stop, structured_outputs, tool_choice, tools, and verbosity. Tools and tool_choice let you wire up function calling, structured_outputs and response_format constrain the shape of the reply, and the reasoning parameters control how much deliberate thinking the model applies.

One practical consideration is that this is the batch variant, which generally implies asynchronous or deferred processing rather than the lowest-latency interactive path; if you need real-time turn-by-turn responses, the non-batch version of the same model is likely the better fit. For workloads that do not need Opus-tier capability, a smaller Claude model may be adequate. Also note that output is text only, so tasks requiring generated images or audio fall outside this model's scope.