Claude Sonnet 4.5 is Anthropic’s Sonnet edition focused on software development, complex agent tasks and computer use, with improvements in reasoning and mathematics. It accepts text and images and generates text. The release describes sustained work across multiple steps, code comprehension and changes that involve a broader project rather than a single isolated function. A useful brief identifies the required behaviour, the relevant source material and the artifact to deliver: a proposed patch, an implementation explanation, a document analysis or a structured answer.
For software work, provide the project’s constraints, the relevant files or excerpts, the observed error and a concrete example of the expected result. Ask for the affected code and an explanation of how the proposed change addresses that example. The producer presents the model as suitable for building agents and working through development tasks with tools. The surrounding application reads and edits files, runs a command or check and returns its actual output; the model reasons over that evidence. Keep the proposed change and the completed check distinct in the final report.
Computer use is a separate application of its visual and agent capabilities. Anthropic’s release demonstrates browser interaction, navigating pages and working with spreadsheets through a connected environment. A visual brief can supply a screenshot, identify the goal and define which actions the application permits. The application then performs the selected interaction and returns the changed screen or other result. Similarly, the model can prepare content and code for documents, spreadsheets and slides, while the environment produces the finished file. Text-and-image understanding does not make the model a native image or audio generator.
The exact producer specification lists a 200K-token context window and up to 64K output tokens. Its reliable knowledge cutoff is January 2025, distinct from the later training-data cutoff listed on the page. Keep document boundaries, page labels, units and code identifiers clear in a long input, and reserve space for the requested answer and any reasoning. The producer also documents context awareness for this edition, so its supported workflow tracks the space remaining during a conversation. That helps manage a sustained task within its window; it does not replace the window with a later Sonnet edition’s larger capacity.
Sonnet 4.5 supports ordinary responses and manual extended thinking. In the latter mode, budget_tokens sets a target for the internal reasoning before the visible answer. The documented minimum is 1,024 tokens, and the budget normally must be lower than max_tokens because thinking and the final answer share the generation allowance. The budget is a target rather than a promise to consume exactly that many tokens. This edition does not have the adaptive-thinking mode or effort controls described for later Sonnet models. Choose the requested analysis depth and leave enough allowance for a usable final answer.
The producer’s interleaved-thinking beta lets Sonnet 4.5 reason between tool calls within a turn, using the returned observations to decide its next step. It requires the documented beta feature in the supported Messages workflow. In this case, the thinking budget can exceed max_tokens because it spans multiple thinking blocks in the same assistant turn. Return tool results with the associated thinking content unmodified, including its signature; editing that content invalidates it. A research or debugging task can therefore alternate a proposed check, the real returned result and an updated analysis without treating a textual proposal as a completed action.
For visual material, the producer supports JPEG, PNG, GIF and WebP and uses only the first frame of an animation. The standard native image tier used by this edition has a 1,568-pixel long-edge limit and a 1,568-visual-token limit; larger images are downscaled. The producer’s image guide lists up to 100 images per API request for 200K-context models, with image and overall request-size conditions that can be reached first. Put images before the question when practical, label multiple images and preserve small text or chart labels. The guide recommends clear images and retaining enough surrounding context to interpret the visible detail.
Context editing and a memory tool were introduced alongside this model to support longer agent workflows. These are application-supported ways to retain useful state or remove old tool material; they are not an unlimited model memory or the later server-compaction feature. For a multi-stage task, preserve the decisions already made, the outstanding question and the evidence needed for the next step. Anthropic also describes stronger resistance to hidden instructions in external content, while noting that model-specific classifiers can sometimes flag benign material. The useful result remains a reviewable artifact tied to supplied evidence, with the exact Sonnet 4.5 modes and limits kept separate from later releases.
", "aboutText": "Claude Sonnet 4.5 is Anthropic’s Sonnet edition focused on software development, complex agent tasks and computer use, with improvements in reasoning and mathematics. It accepts text and images and generates text. The release describes sustained work across multiple steps, code comprehension and changes that involve a broader project rather than a single isolated function. A useful brief identifies the required behaviour, the relevant source material and the artifact to deliver: a proposed patch, an implementation explanation, a document analysis or a structured answer.\n\nFor software work, provide the project’s constraints, the relevant files or excerpts, the observed error and a concrete example of the expected result. Ask for the affected code and an explanation of how the proposed change addresses that example. The producer presents the model as suitable for building agents and working through development tasks with tools. The surrounding application reads and edits files, runs a command or check and returns its actual output; the model reasons over that evidence. Keep the proposed change and the completed check distinct in the final report.\n\nComputer use is a separate application of its visual and agent capabilities. Anthropic’s release demonstrates browser interaction, navigating pages and working with spreadsheets through a connected environment. A visual brief can supply a screenshot, identify the goal and define which actions the application permits. The application then performs the selected interaction and returns the changed screen or other result. Similarly, the model can prepare content and code for documents, spreadsheets and slides, while the environment produces the finished file. Text-and-image understanding does not make the model a native image or audio generator.\n\nThe exact producer specification lists a 200K-token context window and up to 64K output tokens. Its reliable knowledge cutoff is January 2025, distinct from the later training-data cutoff listed on the page. Keep document boundaries, page labels, units and code identifiers clear in a long input, and reserve space for the requested answer and any reasoning. The producer also documents context awareness for this edition, so its supported workflow tracks the space remaining during a conversation. That helps manage a sustained task within its window; it does not replace the window with a later Sonnet edition’s larger capacity.\n\nSonnet 4.5 supports ordinary responses and manual extended thinking. In the latter mode, budget_tokens sets a target for the internal reasoning before the visible answer. The documented minimum is 1,024 tokens, and the budget normally must be lower than max_tokens because thinking and the final answer share the generation allowance. The budget is a target rather than a promise to consume exactly that many tokens. This edition does not have the adaptive-thinking mode or effort controls described for later Sonnet models. Choose the requested analysis depth and leave enough allowance for a usable final answer.\n\nThe producer’s interleaved-thinking beta lets Sonnet 4.5 reason between tool calls within a turn, using the returned observations to decide its next step. It requires the documented beta feature in the supported Messages workflow. In this case, the thinking budget can exceed max_tokens because it spans multiple thinking blocks in the same assistant turn. Return tool results with the associated thinking content unmodified, including its signature; editing that content invalidates it. A research or debugging task can therefore alternate a proposed check, the real returned result and an updated analysis without treating a textual proposal as a completed action.\n\nFor visual material, the producer supports JPEG, PNG, GIF and WebP and uses only the first frame of an animation. The standard native image tier used by this edition has a 1,568-pixel long-edge limit and a 1,568-visual-token limit; larger images are downscaled. The producer’s image guide lists up to 100 images per API request for 200K-context models, with image and overall request-size conditions that can be reached first. Put images before the question when practical, label multiple images and preserve small text or chart labels. The guide recommends clear images and retaining enough surrounding context to interpret the visible detail.\n\nContext editing and a memory tool were introduced alongside this model to support longer agent workflows. These are application-supported ways to retain useful state or remove old tool material; they are not an unlimited model memory or the later server-compaction feature. For a multi-stage task, preserve the decisions already made, the outstanding question and the evidence needed for the next step. Anthropic also describes stronger resistance to hidden instructions in external content, while noting that model-specific classifiers can sometimes flag benign material. The useful result remains a reviewable artifact tied to supplied evidence, with the exact Sonnet 4.5 modes and limits kept separate from later releases.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "4cb3b1638fb4f984c1ec2b1c29d922851a435041f7d0facaa44d471b51a1d53a", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "claude-sonnet-4-5-20250929-thinkingClaude Sonnet 4.5 is Anthropic’s text-and-image model for software development, complex agents, computer use and analytical work. It supports direct responses or manual extended thinking, with reasoning between tool calls in the documented beta workflow.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Claude Sonnet 4.5 is Anthropic’s Sonnet edition focused on software development, complex agent tasks and computer use, with improvements in reasoning and mathematics. It accepts text and images and generates text. The release describes sustained work across multiple steps, code comprehension and changes that involve a broader project rather than a single isolated function. A useful brief identifies the required behaviour, the relevant source material and the artifact to deliver: a proposed patch, an implementation explanation, a document analysis or a structured answer.
For software work, provide the project’s constraints, the relevant files or excerpts, the observed error and a concrete example of the expected result. Ask for the affected code and an explanation of how the proposed change addresses that example. The producer presents the model as suitable for building agents and working through development tasks with tools. The surrounding application reads and edits files, runs a command or check and returns its actual output; the model reasons over that evidence. Keep the proposed change and the completed check distinct in the final report.
Computer use is a separate application of its visual and agent capabilities. Anthropic’s release demonstrates browser interaction, navigating pages and working with spreadsheets through a connected environment. A visual brief can supply a screenshot, identify the goal and define which actions the application permits. The application then performs the selected interaction and returns the changed screen or other result. Similarly, the model can prepare content and code for documents, spreadsheets and slides, while the environment produces the finished file. Text-and-image understanding does not make the model a native image or audio generator.
The exact producer specification lists a 200K-token context window and up to 64K output tokens. Its reliable knowledge cutoff is January 2025, distinct from the later training-data cutoff listed on the page. Keep document boundaries, page labels, units and code identifiers clear in a long input, and reserve space for the requested answer and any reasoning. The producer also documents context awareness for this edition, so its supported workflow tracks the space remaining during a conversation. That helps manage a sustained task within its window; it does not replace the window with a later Sonnet edition’s larger capacity.
Sonnet 4.5 supports ordinary responses and manual extended thinking. In the latter mode, budget_tokens sets a target for the internal reasoning before the visible answer. The documented minimum is 1,024 tokens, and the budget normally must be lower than max_tokens because thinking and the final answer share the generation allowance. The budget is a target rather than a promise to consume exactly that many tokens. This edition does not have the adaptive-thinking mode or effort controls described for later Sonnet models. Choose the requested analysis depth and leave enough allowance for a usable final answer.
The producer’s interleaved-thinking beta lets Sonnet 4.5 reason between tool calls within a turn, using the returned observations to decide its next step. It requires the documented beta feature in the supported Messages workflow. In this case, the thinking budget can exceed max_tokens because it spans multiple thinking blocks in the same assistant turn. Return tool results with the associated thinking content unmodified, including its signature; editing that content invalidates it. A research or debugging task can therefore alternate a proposed check, the real returned result and an updated analysis without treating a textual proposal as a completed action.
For visual material, the producer supports JPEG, PNG, GIF and WebP and uses only the first frame of an animation. The standard native image tier used by this edition has a 1,568-pixel long-edge limit and a 1,568-visual-token limit; larger images are downscaled. The producer’s image guide lists up to 100 images per API request for 200K-context models, with image and overall request-size conditions that can be reached first. Put images before the question when practical, label multiple images and preserve small text or chart labels. The guide recommends clear images and retaining enough surrounding context to interpret the visible detail.
Context editing and a memory tool were introduced alongside this model to support longer agent workflows. These are application-supported ways to retain useful state or remove old tool material; they are not an unlimited model memory or the later server-compaction feature. For a multi-stage task, preserve the decisions already made, the outstanding question and the evidence needed for the next step. Anthropic also describes stronger resistance to hidden instructions in external content, while noting that model-specific classifiers can sometimes flag benign material. The useful result remains a reviewable artifact tied to supplied evidence, with the exact Sonnet 4.5 modes and limits kept separate from later releases.
POST /v1/chat/completionstexttext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits claude-sonnet-4-5-20250929-thinking | Default— |
Fieldmessages | Typemessage[] | RequirementRequired | Values & limits Minimum items: 1 | Default— |
Fieldmessages[].role | Typestring | RequirementRequired | Values & limits Model-specific | Default— |
Fieldmessages[].content | Typestring | RequirementOptional | Values & limits Applies when: Text conversation | Default— |
Fieldstream | Typeboolean | RequirementOptional | Values & limits | Defaultfalse |
Text · UTF8 text
messages[].content
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, messages |
Parametermaterial_types | Contracttext |
Parametermessages | Contractmessage array with role/content shape |
Parameterfile_material | Contractactual UTF-8 document inserted as literal text; no file ID inferred |
Parameteroptional_controls | Contractstream |
Parametercontrol_scope | ContractText conversation; stream is a boolean and defaults to false. |
Parameteroutput_modalities | Contracttext |
Parameterhosted_operations | Contractfalse |
Parametermultimodal_input_binding | Contractnull |
Method aliases: /chat/completions
Lifecycle: synchronous_or_streamed
Send exact model ID with messages.
Context length is not an output cap.
A text file may supply literal text; image/audio/file-object inputs need exact binding.
Tariff references do not add request options. Supported parameters are listed for the selected method.
| Configuration | Rate schedule | Charge |
|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Charge Input · Input tokens $0.006 / 1K tokens Output · Output tokens $0.03 / 1K tokens Input · Cache read $0.0006 / 1K tokens Input · Cache write $0.0075 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model":"claude-sonnet-4-5-20250929-thinking","messages":[{"role":"user","content":"hello"}],"stream":false}')\n\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/chat/completions \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("choices",[]):\n if x.get("message",{}).get("content"): print(x["message"]["content"])\nPY_RESULT\nfrom openai import OpenAI\nimport os\n\nif not os.environ.get("AILOOT_API_KEY", "").strip(): raise SystemExit("Set AILOOT_API_KEY before making a request")\nclient=OpenAI(base_url="https://api.ailoot.ai/v1",api_key=os.environ["AILOOT_API_KEY"],max_retries=0)\nresult=client.chat.completions.create(model="claude-sonnet-4-5-20250929-thinking", messages=[{"role":"user","content":"hello"}], stream=False)\nimport base64, json\nfrom pathlib import Path\n\ndef save_result(data, stem="result"):\n # Keep every field, including usage, IDs, reasoning and tool requests.\n Path(stem + ".json").write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")\n print(json.dumps(data, ensure_ascii=False, indent=2))\n def save_blob(encoded, suffix, label):\n raw = base64.b64decode(encoded, altchars=b"-_", validate=True)\n ext = "png" if raw.startswith(b"\\x89PNG\\r\\n\\x1a\\n") else "jpg" if raw.startswith(b"\\xff\\xd8") else "webp" if raw[:4] == b"RIFF" and raw[8:12] == b"WEBP" else suffix\n file = Path(stem + "-" + label + "." + ext)\n file.write_bytes(raw)\n print("Saved", file)\n for index, block in enumerate(data.get("content") or []):\n print("Content", index, json.dumps(block, ensure_ascii=False))\n for choice in data.get("choices") or []:\n print("Choice", choice.get("index"), json.dumps(choice.get("message", choice.get("delta", {})), ensure_ascii=False))\n for index, item in enumerate(data.get("output") or []):\n print("Output", index, json.dumps(item, ensure_ascii=False))\n if item.get("type") == "image_generation_call" and item.get("result"):\n save_blob(item["result"], item.get("output_format") or "bin", "output-" + str(index))\n for ci, candidate in enumerate(data.get("candidates") or []):\n for pi, part in enumerate((candidate.get("content") or {}).get("parts") or []):\n print("Candidate/part", ci, pi, json.dumps(part, ensure_ascii=False))\n blob = part.get("inlineData") or part.get("inline_data")\n if blob and blob.get("data"):\n mime = blob.get("mimeType") or blob.get("mime_type")\n suffix = {"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"}.get(mime, "bin")\n save_blob(blob["data"], suffix, "candidate-" + str(ci) + "-part-" + str(pi))\n\nsave_result(result.to_dict(mode="json"))\njson or SSE
Keep every complete returned response, file and content part.
Keep every choice and its content/tool calls/reasoning. In SSE assemble deltas by choice/tool index, retain the final finish reason and usage.
Keep request/response IDs and headers. No generic text job retrieval URL is supplied.
| Field | Type |
|---|---|
Fieldchoices[].index | Typeinteger |
Fieldchoices[].message.content | Typestring|null |
Fieldchoices[].message.tool_calls | Typetool_call[] |
Fieldchoices[].message.reasoning_content | Typestring |
Fieldchoices[].message.reasoning_details | Typeobject|array |
Fieldchoices[].finish_reason | Typestring|null |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot Ultra claude-sonnet-4-5-20250929-thinking
/chat/completions