Grok 4.7 vs Claude Opus 5 vs GPT-6 Astra vs Gemini 3.8 Flash: What the Launch Numbers Don't Tell You

Four frontier models launched within seven weeks of each other: GPT-6 Astra on September 3, Gemini 3.8 Flash on September 2, and Grok 4.7 on September 21, all landing after Claude Opus 5's July 24 debut. Every vendor published an intelligence score at launch. Since then the independent tracker Artificial Analysis has moved to version 5.0 of its intelligence index. Re-read on 29 October 2026, that version puts Astra at 57, Grok at 34 and the Gemini model at 31. Opus 5 has no v5.0 score yet.

That spread is now wide, and the pricing behind it misleads in a new way. Astra costs 13 times more per output token than the Gemini model, yet on the tracker's own test run it cost less per completed task, because the cheaper models wrote far more tokens to get there. That gap between the rate card and what a task actually costs is the story every launch post skipped.

Grok 4.7's scorecard on the tracker as it read on 22 Sept 2026, under the previous index version. A month later the same page showed 34 on index v5.0 and 106 tokens/sec.

This guide is for a team choosing an API model for production work now, not for whoever wins next month's cycle. Short version: the Gemini model is the value pick for short, high-volume calls, Opus 5 (or its successor) is the one to reach for when a coding agent works unsupervised, Astra now holds a clear reasoning lead, and Grok 4.7's real advantage was reach, not intelligence.

What Changed Since 8 October

Re-checked against each model's page on the tracker, the Gemini API rate card and the Opus 5 system card:

• The tracker re-scored the field on index v5.0, so the older 53/51/46/41 figures are gone.
• Claude Opus 5.5 shipped on 22 September and scores 58 on v5.0, at a lower rate than Opus 5.
• Gemini 4 Argon is now on sale in the paid API, so the earlier "you cannot buy it yet" note no longer holds.
• The Gemini model's launch rate ends on 31 December.

The four picks below hold, with Opus 5.5 named as the newer Anthropic option.

The Verdict: Who Should Pick Which Model

In one pass, before the detail:

• High-volume, latency-sensitive work (drafting, extraction, classification): the Gemini model, for the per-token price and the speed.
• An autonomous coding agent working unsupervised: Opus 5, for its coding score, or Opus 5.5 for a new build.
• Research-grade reasoning or a contractual no-retention requirement: Astra, which now leads the composite by a wide margin.
• A drop-in swap for a team already on Cursor or Grok Build: Grok, for the reach rather than the benchmarks.

Pick Gemini 3.8 Flash if you are running high-volume, latency-sensitive work: support drafts, data extraction, first-pass classification. Generating at 143 tokens per second on the tracker, it costs a fraction of the alternatives per call. Its index score of 31 is the lowest of the four, so keep it on tasks that do not need long reasoning.

Pick Claude Opus 5 if the job is an autonomous coding agent that has to open pull requests without a human checking every step. It posted 96.0% on SWE-bench Verified in its own system card, the strongest coding figure of the four. For a new build, weigh the cheaper Opus 5.5 first.

Pick GPT-6 Astra when the task genuinely needs reasoning headroom: novel scientific reasoning, long agentic chains where small errors compound, or work where a contractually guaranteed no-retention mode matters. It has the highest rate card of the four, but its 57 on index v5.0 is a 23-point lead over Grok, not a narrow edge.

Grok 4.7 is the pick for reach rather than raw capability. It shipped day one inside Cursor and its own coding tool alongside the standard API, at $2/$6 per million tokens, according to xAI's launch materials as covered by Kingy AI. If your team already lives inside those tools, that convenience is real. Choosing an API model from a blank slate, its 34 score and the coding numbers below put it behind the two leaders.

Opus 5's scorecard as captured on 22 Sept 2026. The price and 1M token window still matched a month later; the index score shown is from the previous version.

Grok 4.7 vs Claude Opus 5, Side by Side

The short answer: Opus 5 is the stronger coding model and Grok 4.7 is the cheaper, faster one per token. Pick the first when an agent writes code unsupervised, and the second when per-call price and editor access matter more than benchmark depth.

Figures from each model's page on the tracker, read on 29 October 2026.

The coding gap is the deciding row, and it is measured on two different tests. Opus 5 scored 96.0% on SWE-bench Verified in its system card. Grok scored 26% on Terminal-Bench 4.0 in The Decoder's launch coverage, against 55% for Claude Fable 5.1 on the same test. No shared benchmark covers both models in the sources checked, so read these as two signals pointing the same way, not as one ranking.

Long prompts narrow that price lead. The xAI rate doubles above a 200,000-token prompt, according to CellCog's launch coverage, and its window stops at 500,000 tokens. An agent that feeds a whole repository into one prompt pays more than the headline rate suggests, and may not fit at all.

Grok 4.7 vs GPT-6 Astra, Side by Side

The short answer: Astra is far stronger, scoring 57 to Grok's 34 on index v5.0, and despite a rate card five times higher on input it cost less per completed index task. The xAI model wins on speed and on short calls where per-token price dominates.

Same source and date as the table above. Astra's context window is 1M tokens against 500K.

The per-task row is the surprise. Grok wrote 140 million output tokens to work through the tracker's evaluation set and Astra 54 million, so the cheaper tokens bought less finished work per dollar on that test. That holds for reasoning-heavy tasks; for a one-paragraph classification call, the per-token rate still decides the bill.

On coding, Astra reached 57.7% on Terminal-Bench 4.0 against 26% for its rival, per The Decoder's benchmark coverage. That is one shared benchmark, which makes this pair easier to compare than Grok against Opus.

What Each Model Actually Costs

Vendors quote input and output rates separately, and the gap between them is where buyers get surprised. Figures below are per million tokens, current as of the dates shown.

Rates as listed on the tracker on 29 October 2026.

Google's Gemini API pricing page lists the Gemini rate as valid through 31 December 2026, rising to $1.50 input and $7.50 output from 1 January 2027. Budget a contract that runs into next year at the higher figure. The xAI model's window is the odd one out at 500,000 tokens against roughly 1 million for the other three.

A worked example makes the output-token gap concrete. A support team processing 1,000 tickets a day, averaging 500 input tokens and 200 output tokens per ticket, spends roughly this much per day at each vendor's rates:

That is a $34 line item against a $450 one for identical ticket volume, and the cheapest line doubles in January. The gap only closes if the cheaper answers need enough human correction to erase the savings, a real risk beyond routine drafting.

Per-Token Price Is Not Per-Task Cost

The tracker now measures these models on the nine evaluations of index v5.0. On that version GPT-6 Astra scores 57, Grok 4.7 scores 34, and Gemini 3.8 Flash scores 31. Claude Opus 5 has not been run on v5.0; its successor scores 58.

The tracker also publishes what each model cost to work through that evaluation set, and it reorders the field:

Per-task costs are the tracker's own weighted averages.

The reason is verbosity. The Gemini model wrote 130 million output tokens on the evaluation set and Grok 140 million, against Astra's 54 million. On hard, multi-step tasks the cheap rate is spent many times over. On short tasks it is not, which is why the Gemini model still wins the support-ticket example above.

Anthropic's Fable 5.1 shows how little a single coding test tells you. In the benchmark round covered by The Decoder it scored 55% on Terminal-Bench 4.0 against 26% for Grok. It is not a budget option, though: the tracker lists it at $10/$50 per million tokens, the same rate card as Astra.

Where Coding Ability Actually Diverges

The composite score is a blend. For engineering teams, the coding-specific numbers matter more, and here the benchmark version matters as much as the result. Terminal-Bench 4.0 is a newer, harder revision than 2.1, so scores across the two do not compare directly.

Source: The Decoder's benchmark coverage, accessed September 22, 2026, and the Opus 5 system card, re-read for this update. Opus 5 was not run against version 4.0 in the sources checked for this guide, so it is absent from that column. SWE-bench Verified and Terminal-Bench measure related but different things, patch accuracy against a fixed test suite versus open-ended terminal task completion, so read the two rows as complementary rather than one ranking.

The practical read: if your coding workload looks like SWE-bench, resolving a defined issue against a known repo, that 96.0% figure is the strongest signal available right now. If it looks like Terminal-Bench, open-ended terminal and tool use, the 57.7% score currently leads the field on that exact test.

Where Grok 4.7 Wins: Reach, Not Intelligence

Grok's real story at launch was not the benchmark table. It shipped with same-day access through two coding tools and the standard API, according to launch coverage from Kingy AI, matching what the vendor itself published. That is a faster path into an existing coding workflow than waiting on a separate integration.

One claimed advantage does not hold up under a second source. Musk said SpaceX engineering data was folded into training, but a CellCog review of the actual launch post found no mention of that dataset in the vendor's own materials. This guide treats the claim as unconfirmed rather than repeating it as fact.

What is confirmed is pricing that matched the prior release exactly, a same-day rollout into tools developers already use, and output speed second only to the Gemini model here. For a team already paying for those seats and needing a drop-in swap with no new procurement step, that convenience has real value. Our earlier coverage of the coding distribution deal goes deeper on what changed there. Just do not confuse fast reach with strong benchmarks.

Where Claude Opus 5 Wins: Coding You Can Trust Unsupervised

SWE-bench Verified is built to resist gaming: a fixed set of real GitHub issues with a known correct patch, penalizing a model that claims success without resolving the issue. Opus 5's 96.0% there, an average over five trials in Anthropic's own system card, is the strongest signal in this comparison for one specific job: an autonomous agent that opens a pull request without a human reviewing every intermediate step.

Its million-token context window also matches Astra and the Gemini model rather than trailing the smaller xAI window, which matters for an agent that needs a whole repository in view. Opus 5 remains on sale, but its successor now costs less, so a team starting today should test both.

For how this class of model stacks up against a dedicated coding-agent product, our Cursor Composer vs Claude Code piece is the closer look, and Hokai's coding-assistant category covers the wider field of products built on top of models like this one.

Where GPT-6 Astra Wins: The Extra Reasoning Headroom

Astra's composite score of 57 is the highest of the four on index v5.0, and its 57.7% leads the Terminal-Bench 4.0 column of this guide. OpenAI also offers eligible enterprise customers a mode that deletes inputs and outputs the moment a request finishes, with no staff review, rather than the 30-day default most of the field uses.

The case against it is the rate card: $50 per million output tokens is double Opus 5 and over eight times Grok. That premium shows up on short calls, where every token is billed and little reasoning is needed. On long reasoning chains and agentic runs it largely disappears, because Astra uses far fewer tokens to finish, which is exactly where its 23-point lead over Grok decides the outcome.

Where Gemini 3.8 Flash Wins: Price and Speed, Not a Close Second

Gemini 3.8 Flash generates at 143 tokens per second, ahead of Grok 4.7's 106 and nearly double GPT-6 Astra's 76, per the same tracker. Paired with its per-token pricing, reach for it when the job is high-volume and latency-sensitive rather than reasoning-bound: drafting, extraction, summarization, anything run at scale where a score of 31 is enough headroom.

For scale, Google's own previous release, Gemini 3.6 Flash, has since been deprecated in favor of newer Flash releases, which shows how fast this specific product line moves.

Outside this four-way comparison it no longer sets the price floor. Claude Haiku 5.5 scores 36 on index v5.0, above the Gemini model's 31, at $0.10 input and $0.50 output per million tokens on the tracker. Haiku 5.5 is priced for prompts up to 100,000 tokens and steps up above that, so check your prompt sizes before switching.

Two models from the previous version of this list no longer undercut it. Mistral Large 4 Preview is back at $1.36/$4.18, above the Gemini rate, and scores 25; Mistral AI's half-price sale has ended. Ling 3.1 Flash has no v5.0 score yet. Hokai's cheapest LLM API guide tracks those rate cards side by side, and what Mistral Large 4 is for covers the preview.

The Fine Print: These Are the Top-Effort Scores

Every number above is the highest reasoning-effort configuration each vendor publishes, not necessarily what a default API call returns. The tracker labels its test runs accordingly: Grok's entry is marked "xhigh," Astra's is marked "max," and the Gemini model's is marked "high." Opus 5's listing specifies "Max" effort, and its successor's run used 220 million output tokens on the index, well above the 64 million median.

That matters for two reasons. First, a lower effort setting on any of the four will typically score below the numbers in this guide and cost less per response, since more reasoning tokens are what the higher tiers spend to earn their extra points. Second, it means the price comparisons above are apples to apples: all four are quoted at the configuration that produced the composite score being compared, not a cheaper default that would understate the real cost of matching that score.

For most production workloads, that top tier is overkill. A support-ticket classifier does not need the priciest model's maximum setting any more than a spreadsheet needs a supercomputer. Test each vendor's default and low-effort tiers against your own task before assuming the number in this guide is the price you will actually pay.

How Each Model Handles Your Data

All four vendors say API inputs are not used to train their models by default, but the retention terms behind that promise differ in ways that matter for regulated workloads.

Anthropic classifies Opus 5 under the EU AI Act as a general-purpose system with systemic-risk obligations, and offers both US and EU residency. Google's exception matters for the free tier specifically: users in the EEA, Switzerland and the UK get the paid-tier terms applied even on unpaid usage.

If your workload cannot tolerate any server-side retention, the two priciest models here currently offer the cleanest enterprise paths. Grok's zero-retention setting costs you the stateful features that make an agent workflow convenient in the first place.

The Turn: Maybe the Composite Score Doesn't Matter Here

The strongest objection to everything above is that a nine-benchmark blend was never built to predict performance on your specific task. A model scoring five points lower on a broad average can still be the better choice if its individual strengths line up with what you are building, and a five-point average gap can come from one or two tests that have nothing to do with your use case.

That objection is fair, and it is exactly why this guide breaks the comparison into coding, price, speed and data handling separately rather than stopping at one number. The composite is a useful first filter for narrowing four models to two worth testing. It is a poor substitute for running your own eval set against the finalists.

This refresh is the proof of a second, quieter problem: a tracker's score is tied to its suite version. Within five weeks every model here was re-scored on a new suite, and Grok fell from 46 to 34 without the model changing. Different evaluation services also weight benchmarks differently and update on their own schedules.

Treat any single number here as a snapshot from one respected source, not a settled fact about which model is smarter in some absolute sense. The benchmark tables further up this guide, sourced independently and dated, are the more durable evidence. A single composite score is only the quick filter that gets you to those tables, not the final word.

Switching Cost: What Moving Between These Four Actually Takes

All four sit behind a standard REST API with OpenAI-compatible or vendor-native SDKs, so the code-level switching cost is genuinely low for a simple completion call. It rises fast once a workflow depends on vendor-specific features.

Moving off the xAI zero-retention setting means giving up its stateful file, collection and batch features, which can mean rewriting an agent's state-management layer. Moving between the other vendor families means re-tuning prompts: they respond differently to system-prompt structure and tool-call formatting, and a prompt tuned for one rarely transfers cleanly to another.

The context-window gap matters here too. An agent built around a million-token window, feeding it a whole repository or document set at once, cannot drop into the smallest context window here without adding a retrieval or chunking step first. That is an afternoon of engineering for a small workflow and a real re-architecture for a large one. If you are choosing between models under a specific constraint, open weights or the lowest possible latency among them, Hokai's model recommender walks through that decision by constraint rather than by brand name.

Since Publication: Gemini 4 Argon

Google announced Argon on 30 September 2026, after this comparison was first written. On index v5.0 the tracker scores it 50 with high reasoning, below Astra's 57 and above Grok's 34, at 82 tokens per second and $7.68 per index task.

It is now on sale in the paid Gemini API, with no free tier. Per million tokens:

Rates from the same Google rate card. The lower rate is temporary, so weigh the later one.

Argon does not change the four picks above. It trails Astra on the composite, and from December its $4/$20 rate matches Opus 5.5, which scores higher. It is the model to test if you are already on that API and need more reasoning than the Flash model gives. The live model leaderboard tracks it with the rest of the field.

What Would Change This Verdict

This comparison is a snapshot of four launches inside a seven-week window, and the numbers behind it move. If the priciest model's rate drops toward the mid-pack price, as prior generations from the same vendor have within months of launch, its lead in reasoning stops being a premium purchase and becomes the default pick outright.

If the tracker publishes a v5.0 score for Opus 5, or a Terminal-Bench 4.0 result for either Anthropic flagship lands below the current 57.7% leader, the coding verdict above needs revisiting. And if the xAI model closes that same gap the way the Cursor coding integration covered earlier already closed the distribution gap, its price-to-capability ratio gets a lot more interesting than it is today.

For now, the fastest way to see how these four, and everything else Hokai tracks, stack up on the metric you actually care about is the live model leaderboard, which updates as new scores land rather than staying frozen at launch-day numbers the way this guide necessarily is. If you are unsure which axis matters most for your project, Hokai's Smart Match asks about your actual workload and narrows the field from there instead of asking you to read four spec sheets first.

Frequently asked questions

Is Grok 4.7 better than Claude Opus 5?

For coding, no. Anthropic's system card puts Opus 5 at 96.0% on SWE-bench Verified, while the xAI model managed 26% on the harder Terminal-Bench 4.0. Grok is cheaper at $2/$6 per million tokens against $5/$25, and faster at 106 tokens per second against 59 on Artificial Analysis (29 October 2026).

Is Grok 4.7 good at coding?

It trails the strongest models here on the coding benchmarks checked for this guide: 26% on Terminal-Bench 4.0, compared with GPT-6 Astra's 57.7% and Anthropic's Claude Fable 5.1 at 55% on the same test. Its real strength at launch was distribution, shipping the same day inside Cursor and Grok Build, not raw coding capability.

Which of these four is cheapest to run?

Gemini 3.8 Flash per token, at $0.75 input and $3.75 output per million until 31 December 2026, then $1.50 and $7.50 from 1 January 2027, per Google's pricing page. On Artificial Analysis's reasoning-heavy evaluation set, though, GPT-6 Astra cost less per completed task ($6.24 against $7.77).

Do Grok 4.7, Claude Opus 5, GPT-6 Astra and Gemini 3.8 Flash train on my data?

None of the four train on API inputs by default, according to each vendor's published terms. Grok 4.7 retains encrypted API traffic for 30 days for abuse monitoring before deleting it; GPT-6 Astra and Claude Opus 5 both offer Zero Data Retention to eligible enterprise customers; Gemini 3.8 Flash's paid tier does not train on inputs, though its free tier may.

Which model should I actually use for an autonomous coding agent?

Claude Opus 5. Its 96.0% on SWE-bench Verified, five-trial mean reported by Anthropic, is the strongest published result among the four for resolving real repository issues. For a new build, also test Claude Opus 5.5, which costs less at $4/$20 per million tokens and scores 58 on Artificial Analysis index v5.0.

Covered in this guide

• Grok 4.7: xAI's September 2026 flagship model, built for coding and agentic work with a 500K token context window and a 46.3% CursorBench score.
• Claude Opus 5: Anthropic's July 2026 Opus with a 1M token context and xhigh reasoning effort, still active but superseded by Claude Opus 5.5 on September 22, 2026.
• Gemini 3.8 Flash: Google DeepMind shipped Gemini 3.8 Flash on September 2, 2026, a multimodal Gemini 3 model tuned for long-horizon coding and agentic enterprise work.
• GPT-6 Astra: OpenAI's flagship model, launched September 2026 as the first ever rated at the Preparedness Framework's Critical cybersecurity level.
• Anthropic: Anthropic is a Public Benefit Corporation started by 7 ex-OpenAI researchers. It builds the Claude model family and counts Amazon and Google among its largest backers.
• Claude Fable 5.1: Anthropic's second Mythos-class model, released Sept 1, 2026, for long-horizon agentic coding and research at Fable 5's per-token rates.
• Claude Haiku 5.5: Anthropic's smallest and fastest 5.5 model, released October 7, 2026, with a 1M context window, built for high-volume work.
• Claude Opus 5.5: Anthropic's September 2026 flagship LLM with a 1 million token context window, built for long-running agentic coding and computer use.
• Cursor: Cursor is an AI code editor built on VS Code, widely adopted across large enterprise engineering teams, with Agent Mode, Tab completion, and Cloud Agents.
• Gemini 3.6 Flash: Released July 21, 2026, Gemini 3.6 Flash is Google DeepMind's Flash-tier model built for agentic coding, computer use and long-context work.
• Gemini 4 Argon: Google DeepMind's Gemini 4 frontier model, generally available on the paid Gemini API: 1M-token context and output, text, image, video, audio and PDF input.
• Google DeepMind: Alphabet's AI research lab, created from DeepMind and Google Brain. It makes the Gemini, Gemma and AlphaFold models; Gemini 4 Argon is on general sale and the newer Carbon is limited to partners.
• Ling 3.1 Flash: InclusionAI's text-only reasoning model for coding agents and long tasks, released on 30 September 2026 as the successor to InclusionAI's July flash model.
• Mistral AI: Paris lab behind Mistral Large 4 (1T parameters, preview Oct 2026) and Vibe, Studio, Forge and Compute. Raised a €3B Series D at over €21B in Sept 2026.
• Mistral Large 3: Mistral Large 3 (Dec 2025) is a 675B MoE model (41B active) with 256K context and Apache 2.0 open weights.
• Mistral Large 4: Mistral Large 4 is Mistral AI's largest open-weight multimodal MoE model, launched 6 Oct 2026 in public preview with image input.
• OpenAI: OpenAI builds ChatGPT, the GPT-6 family (Astra, Sol, Luna), Codex and the OpenAI API. It reports 1.2B weekly ChatGPT users and an $852B valuation after its March 2026 round.
• xAI: Elon Musk's AI company (roughly 4,000 to 4,900 employees) builds the Grok models and merged into SpaceX, taking the combined business public on Nasdaq as the largest IPO on record.

Sources