explainx / blog
The context window is how many tokens a model can condition on in one request—input plus the budget reserved for a reply. Here is a plain definition, how it differs from parameter count, and a comparison table for flagship 2026 models (GPT-5.4, Claude 4.7 family, Gemini 3.1 Pro, Meta Llama 4) with links to the canonical docs.
Apr 22, 2026
Bigger is not a synonym for smarter, but parameter count is still the first axis people use to compare scale. This guide explains what parameters are, how mixture-of-experts changes the math, and which flagship models still publish size—and which do not.
Apr 22, 2026
If you have ever seen
Jun 29, 2026
For the first time, more U.S. businesses on Ramp pay for Anthropic than OpenAI — 34.4% vs 32.3% in April 2026, with Anthropic quadrupling share in a year while OpenAI grew 0.3%. The viral X thread tied it to loop engineering and Dario Amodei's claim that some Anthropic engineers barely write code anymore. Here is what the index actually counts, what it does not, and why workflow beats chat.
The context window (context length) is the token budget a single request can use: the full prompt you send (system text, user message, prior turns, tool definitions, retrievals, tool results) plus the space allowed for the new completion, up to a max output cap. It is a limit on how much the model can attend to at once, not the same as parameter count or how usage is metered in tokens—though all three appear on the same pricing pages.
If the request does not fit, the system may return an error, truncate older content, or compress it—behavior depends on the client and API.
Three ideas often collide in vendor marketing:
| Concept | Meaning |
|---|---|
| Context window | Max tokens one forward pass can attend to right now |
| Training data | Corpus used during pre-training—not fully visible in any single prompt |
| Product “memory” | Vendor features (projects, compaction, LLM wikis) that persist facts across sessions |
A model “knowing Python” is weights, not context. Pasting your entire repo into chat is context. Confusing the two leads to under-pasting (missing files) or over-pasting (blowing the window on turn three).
Rough planning (English prose):
When in doubt, log usage.prompt_tokens from API responses for a week and build an internal table of “typical task sizes.”
Before shipping an agent, fill in:
| Bucket | Your estimate | Notes |
|---|---|---|
| System + developer instructions | Include tool JSON schemas | |
| Retrieved chunks (RAG) | Top-k × chunk size | |
| Conversation history | Turns × avg turn size | |
| Tool results (last turn) | Often underestimated | |
| Headroom for output | Max completion tokens |
If the sum approaches the model window, plan compaction or external memory before launch—not after users hit truncation errors.
Is context the same as memory? No—memory products persist facts; context is per-request unless you re-inject stored state.
Does a bigger window replace RAG? Rarely. Retrieval scales to corpora larger than any window and keeps citations fresh.
What breaks first on long threads? Usually cost and latency, not the theoretical token cap—monitor both when enabling 1M-class models.
For coding agents specifically, see Claude Code large codebase tips and Headroom compression.
When a blog post says “1M context,” ask:
The table in this guide is a starting point—your contract and region may differ.
Read next: What are tokens? · Parameters explained · Caveman token compression · Claude Opus 4.7
Treat context window as a hard budget in agent design reviews—the same way you cap cloud spend. Teams that skip this step discover truncation only when a long-running support thread silently drops the system prompt.
Even when a model advertises 1M tokens, vendors may charge premium rates above 272k input tokens or route long requests to slower queues. Budget owners should model p95 prompt size × price per million × daily requests before enabling “paste entire repo” workflows in internal tools.
Bottom line: Context window is the ceiling on one request’s working set—design agents assuming you will hit that ceiling weekly, not once a year.
When comparing Claude, GPT-5.4, and Gemini 3.1 Pro for an agent project, build a spreadsheet with your real prompts—not vendor maxima—as the input column.
For coding agents, tool JSON and MCP server definitions often consume tens of thousands of tokens before the user speaks—budget them in the same worksheet as chat history (MCP guide).
Context window = max tokens one request can use (prompt + completion budget). It is not parameter count, not training data, not vendor marketing alone. Size agents with real prompts, tool overhead, and compaction plans before you promise “unlimited chat” to users.
Re-check vendor model pages when upgrading—context and max output limits change independently of model name.
See also Headroom compression when prompts routinely exceed half your window.
Vendor numbers change; confirm on the model page for your API route.
The table below is a 2026 documentation snapshot—always re-check the model page in your environment (direct API, Bedrock, Vertex, Foundry, etc.).
| Model (line) | Context window (tokens) | Max output (typical) | Where to confirm |
|---|---|---|---|
| Claude Opus 4.7 | 1,000,000 (1M) | 128,000 (sync Messages) | Anthropic models overview |
| Claude Sonnet 4.6 | 1,000,000 (1M) | 64,000 | same |
| Claude Haiku 4.5 | 200,000 | 64,000 | same |
| GPT-5.4 (API) | 1M long-context; see guide for 272K vs 1M behavior | up to 128,000 (see model page) | Latest model, GPT-5.4 |
| Gemini 3.1 Pro (Preview) | 1,000,000 input (1M) | on the order of 64k output (varies) | Google AI dev docs, DeepMind |
| Llama 4 Maverick (open) | 1,000,000 (1M) in model card | server-dependent | Meta model card |
| Llama 4 Scout (open) | 10,000,000 (10M) in model card | server-dependent | same |
Caveats: a large on-paper window does not force you to fill it; latency and cost usually grow with effective length. Multimodal inputs (images, video) use additional or separate budgets on many APIs. Hosted marketplaces can impose tighter caps than the base model card.
Standard transformer attention is heavy in memory for long sequences, so providers invest in kernels, sparsity, chunking, and user-facing compaction to make 1M-class products usable. That engineering is a major reason for long-context tiers and pricing rules—not only marketing.
Suppose you call Claude Opus 4.7 with a 1M context window and 128k max output (sync API, per Anthropic docs):
| Component | Tokens (example) |
|---|---|
| System prompt + tools | 8,000 |
| Retrieved RAG chunks (5 × 2k) | 10,000 |
| Prior conversation (20 turns) | 120,000 |
| User message | 2,000 |
| Subtotal input | 140,000 |
| Remaining for output | up to 128,000 (capped by max output, not “860k unused”) |
You rarely “use the whole million.” Effective length drives latency and bill—see LLM tokens for pricing math.
Multi-turn trap: each turn re-sends prior messages (unless the client compacts). A 50-turn support thread can exceed a 200k window even if each message looked small in isolation.
| Situation | Prefer | Why |
|---|---|---|
| Static corpus (docs, wiki) | RAG + small window | Cheaper; fresher index without re-prompting everything |
| Single huge file (contract, log dump) | Long context (if affordable) | Cross-chunk reasoning without retrieval errors |
| Agent with tools | Tools + compaction | Pass handles, not full payloads every turn (MCP) |
| Multimodal (many images) | Check per-modality caps | Vision tokens often separate from text budget |
Headroom context compression and Karpathy LLM wiki pattern are architectural answers when neither raw long context nor naive RAG alone scales.
| Provider | Gotcha |
|---|---|
| Anthropic | 1M window on Opus/Sonnet but sync max output still capped—read Messages API limits |
| OpenAI | GPT-5.4 documents 272k vs 1M behavior tiers for rate limits and pricing |
| Gemini preview models change token accounting for multimodal inputs | |
| Meta (Llama 4) | Model card lists 10M for Scout—your host may impose a lower cap |
| Bedrock / Vertex | Marketplace wrappers often below vendor maximum |
Always size workloads on the route you actually call, not the headline blog number.
When threads overflow, clients may truncate (drop oldest turns) or compact (summarize). Truncation silently loses facts; compaction loses nuance if summaries are poor. Prefer explicit compaction with stored summaries in an LLM wiki over blind tail dropping.
Vendor numbers change; re-check the model and pricing page for the exact route you use.
The context window caps how much text (and often multimodal input) a model can condition on in one request—distinct from parameter count and token billing. 2026 flagships advertise 1M-class windows, but max output, hosting route, and compaction behavior still bound real workflows.
Practical rule: prefer RAG + tools + compaction for unbounded corpora; use long context when you need cross-chunk reasoning on a single artifact and can afford latency and cost. Re-read vendor docs before sizing agent memory or support thread limits.
