xAI’s flagship for this generation, Grok 4.7, is live on unified API catalogs — and it arrives with the sharpest price-to-flagship ratio in the October 2026 lineup: $1.60 per million input tokens and $4.80 per million output, with coding, reasoning, tools and flagship tags on a 500K-token context window.
This is a spec-and-pricing analysis, not a synthetic benchmark. Every figure comes from the live MeshTok catalog on 2026-10-19 and is reproducible on the pricing page. The goal: decide where Grok 4.7 fits in a production stack — and where the rest of the Grok family beats it.
Grok 4.7: the spec sheet
Available today through the MeshTok unified API at https://meshtok.com/v1 under the model id x-ai/grok-4.7:
| Field | Value |
|---|---|
| Vendor | xAI (US) |
| Context window | 500,000 tokens |
| Input price (per 1M) | $1.60 |
| Output price (per 1M) | $4.80 |
| Capabilities | coding, reasoning, tools, flagship |
| Released | 2026-10 |
Three things stand out:
- It is the cheapest
flagship-tagged Western model in the catalog. Among US-lab flagships, only Opus 5.5 ($4.00/$20.00) and Fable 5.1 ($10.00/$50.00) sit above it — both at 2.5× to 10× Grok’s input price. The Chinese flagships (Hunyuan 4 Preview at $0.89/$2.68, DeepSeek V4 Pro at $0.67/$1.34) undercut it on price but Grok carries the Western-lab ecosystem story (tool-calling conventions, agentic SDK assumptions) many stacks are built around. - The output price is the real headline. At $4.80/M output, Grok 4.7 costs less than half of Sonnet-tier rivals ($10.00) — and code generation is output-heavy work. This is the number that moves monthly bills.
- 500K context is the one compromise. Most October-generation reasoning models ship 1M windows (Sonnet 5.5, GPT-6.1 Sol, Claude Opus 5.5). Grok’s 500K still swallows entire repositories for most codebases; only extreme long-document workloads will feel the cap.
The Grok family: which one to pick?
The catalog carries five Grok models spanning a 24× price range:
| Model id | Context | Input $/M | Output $/M | Role |
|---|---|---|---|---|
x-ai/grok-4.7 | 500K | $1.60 | $4.80 | flagship reasoning + coding |
x-ai/grok-4 | 256K | $3.00 | $15.00 | previous-gen flagship (snapshot) |
x-ai/grok-3 | 131K | $3.00 | $15.00 | legacy |
x-ai/grok-4-fast | 2M | $0.20 | $0.50 | speed tier, giant context |
x-ai/grok-4.1-fast | 2M | $0.20 | $0.50 | speed tier, newer snapshot |
Reading the matrix:
- Grok 4.7 vs Grok 4 — the previous flagship costs nearly 2× more on both axes with half the context. Unless you are pinned to the 4 snapshot for eval reproducibility, 4.7 supersedes it on paper: cheaper, smarter per the vendor’s positioning, bigger window.
- Grok 4.1-fast is the sleeper. At $0.20/$0.50 with a two-million-token window, it is the cheapest per-token route to huge-context workloads in the entire catalog — cheaper on input than GPT-6 Luna ($0.10 in but $0.50 out, 1M window) on the output side where long-context runs actually bill. Whole-book analysis, massive log triage, monorepo-wide question answering: this is its job.
- Grok 3 exists for legacy pins only; new work has no reason to start there.
What a month of Grok traffic costs
Reference workload: 1M input + 0.25M output tokens daily, 30 days:
| Model | Daily | Monthly |
|---|---|---|
| Grok 4.7 | $1.60 + $1.20 = $2.80 | $84.00 |
| Grok 4 / Grok 3 | $3.00 + $3.75 = $6.75 | $202.50 |
| Grok 4.1-fast | $0.20 + $0.125 = $0.325 | $9.75 |
For the same workload, Claude Sonnet 5.5 and GPT-6.1 Sol both cost $135.00, Claude Opus 5.5 costs $270.00, and DeepSeek V4 Pro costs $30.00 (see our coding API ranking). Grok 4.7 lands at 62% of the Sol/Sonnet price with flagship tagging — and 31% of Opus’s.
Head-to-head: where Grok 4.7 wins and where it does not
| Rival | Their edge | Grok 4.7’s counter |
|---|---|---|
anthropic/claude-sonnet-5.5 ($2.00/$10.00) | vision tag, 1M window | 43% cheaper monthly ($84 vs $135), flagship tag |
openai/gpt-6.1-sol ($2.00/$10.00) | 1M window, OpenAI ecosystem | 43% cheaper monthly, cheaper output for code gen |
deepseek/deepseek-v4-pro ($0.67/$1.34) | price, vision, 1M | Western-lab tool-calling conventions, flagship branding |
anthropic/claude-opus-5.5 ($4.00/$20.00) | 1M window, vision, stronger positioning | 68% cheaper monthly at the same agentic mission |
The honest summary: Grok 4.7 is the value pick of the $1–2 flagship-adjacent tier. You give up 500K of context and the vision modality; you keep flagship reasoning/coding and cut the bill by 40–70% versus Sonnet 5.5, GPT-6.1 Sol or Opus 5.5. If screenshots drive your workflow, Sonnet 5.5’s vision tag is worth the premium; if your prompts are text-and-tools, the premium is hard to justify.
How to call Grok 4.7
from openai import OpenAI
client = OpenAI(
base_url="https://meshtok.com/v1",
api_key="sk-your-MeshTok-key",
)
resp = client.chat.completions.create(
model="x-ai/grok-4.7", # or grok-4.1-fast for the $0.20 speed tier
messages=[{"role": "user", "content": "Review this architecture."}],
)Prices reflect the MeshTok catalog on 2026-10-19; the live source of truth is the pricing page and the Grok 4.7 model page. For the full competitive field at this tier, see GPT-6.1 Sol vs Claude Sonnet 5.5 vs Grok 4.7.