Claude: long-context judgment and careful generation
Anthropic’s Claude family — a general-purpose LLM for writing, long documents, coding, and assistants. Profiled here so comparisons have a durable URL, not as a tutorial.
What Claude is
Claude is Anthropic’s general-purpose LLM family. It generates text, code, and tool calls. The product story emphasizes careful writing, long context, and safer defaults. That is positioning, not a medal from this site.
Like ChatGPT, Claude is a generator. Structured outputs do not turn it into a decision model. If the output set is a closed enum and the next hop is code, read Jev vs Claude before you pay generation prices for a label.
Developer / company
Anthropic. Consumer surface: claude.ai. Builder surface: the Claude API, plus Amazon Bedrock, Google Vertex AI, and Microsoft Foundry listings.
Launch date and current versions
Claude first shipped in 2023. As of 20 September 2026, Anthropic’s public deprecation table lists active Claude 5 aliases (including Sonnet 5, Opus 5, and newer Fable-class SKUs) alongside still-active 4.x names such as Sonnet 4.6 and Opus 4.8. Original Claude 4 / 4.1 snapshots were retired earlier in 2026. Pin the alias. Do not assume last year’s model id still answers.
Pricing (vendor claim)
Anthropic’s published standard global list as of this desk date:
- Claude Sonnet 5: $2 input / $10 output per 1M tokens. Anthropic says the $2 / $10 introductory rate is now the standard price.
- Claude Opus 5 (and recent Opus 4.8): $5 / $25 per 1M.
- Prompt-cache hits are much cheaper than fresh input. Batch is discounted.
Source: Anthropic pricing. Dated snapshot, not a contract. Consumer Claude Pro / Team seats are a different bill.
Input / output
- Input: text and images on current SKUs; very long contexts on the SKUs that advertise them.
- Output: generated assistant text, code, and tool use. JSON / schema constraints are available — still generation.
- Prompt caching is part of the real bill for repeated system prompts. Vendor claim; we have not metered it.
Main use cases
Long documents, careful drafts, coding help, and judgment-heavy assistants. Teams pick Claude when the reader is a human and the cost of a sloppy paragraph is high. Classification is possible. It is not why you bought Claude.
API availability and integrations
First-party Messages API plus the three major clouds. Slightly less “every indie plugin speaks Claude” than ChatGPT; more than enough for production. Deprecation tables are part of the integration story — budget a migration, not just a key.
Strengths
- Strong default for long-context reading and careful prose — as a product stance, backed by a large public conversation, not by our lab.
- Clear SKU ladder (Haiku / Sonnet / Opus, plus newer names) with published cache pricing.
- Cloud listings reduce “we cannot put the key in this VPC” friction.
Limitations
- Still a generator. Wrong tool if you only needed a typed decision.
- Alias retirement in 2026 already bit teams who pinned Claude 4 snapshots.
- Consumer distribution is smaller than ChatGPT. That matters for “my users already have an account.”
- No AIToolsRatings measured reliability or latency table yet.
When to pick something else
- Default consumer assistant / widest plugin graph → ChatGPT.
- Workspace / Search-native grounding → Gemini.
- High-QPS classify/route → Jev or the routing shortlist.
Latest updates
September 2026: Claude 5 family is the live generation; several Claude 4 ids are retired. Sonnet 5’s $2 / $10 list is the workhorse number we will keep dating. Not a Claude tutorial.
Task fit, not a winner
| Task | Claude | Decision model (e.g. Jev) |
|---|---|---|
| Classification | Structured decision with a bounded label set | Possible, usually via generated text or JSON schema |
| Model routing | Native fit: pick a route, return a typed choice | Possible, but you pay generation cost for a decision |
| AI agents | Use as a judge / router / gate, not as the actor | Use as planner, writer, tool-caller, and explainer |
| RAG | Score, filter, or route retrieved chunks | Synthesize an answer from retrieved context |
| Chat | Not the job. No free-form conversation | Primary job |
| Writing | Not the job. No prose generation | Primary job |
| Coding | Not the job, unless you only need a pass/fail or route | Generate, explain, and iterate on code |