Gemini 4 Neon vs Gemini 4 Carbon
For a multi-step automation, how do November token prices compare with the cost of an entire attempt in Zapier’s test?
01 // VERDICT
The short answer
| At a glance | Gemini 4 Neon | Gemini 4 Carbon |
|---|---|---|
| List price · Introductory through 2026-11-30input / output per 1M tokens | $3 / $15 | $2 / $10 (better) |
| List price · From 2026-12-01input / output per 1M tokens | $3 / $15 (better) | $4 / $20 |
Highlighted values mark the lower rate or stronger measured score in the same documented run. The dates apply to the Argon and Carbon rates. The calculator uses your selected period. · Gemini pricing snapshot (verified 2026-11-19)
EXECUTIVE SUMMARY
Trial Neon on workflows like the ones Zapier tested. At high effort it passed 357 of 657 attempts, against 354 for Carbon. Zapier estimated $0.88 per attempt for Neon and $1.94 for Carbon. A fixed-token budget gives a different November result because it holds output and cache hits constant. Carbon is cheaper on that basis through November 30.
Choose Gemini 4 Neon
Your workflows involve repeated tool calls and you can reproduce the completion rate and lower spend per accepted outcome on your own tasks.
Choose Gemini 4 Carbon
You have validated Carbon on long-running jobs, or you can complete a fixed-token workload before its introductory rates expire.
02 // KEY FACTS
Key facts
- On AutomationBench 1.0.6, 54.3% of Neon attempts and 53.9% of Carbon attempts passed at high effort. Zapier used 657 tasks per setting and checked the resulting state of the simulated business apps.
- Zapier applied list prices to observed token use, including reasoning output and cached input. It assumed $0.15 per million Neon cache hits and used a different Carbon cache-read basis from the leaderboard. The $0.88 and $1.94 figures are unbilled estimates across all attempts.
- For the editable Standard budget, each of 40,000 monthly requests has 12,000 total input tokens (8,000 fresh and 4,000 cached), zero cache writes and 2,500 billed output tokens. These counts are assumptions you can replace with your own usage.
- At those counts Neon costs $2,484 per month. Carbon costs $1,656 through November 30, 2026, and $3,312 from December 1. The monthly totals bill 320 million fresh tokens, 160 million cache hits and 100 million output tokens.
- DeepMind identifies Carbon as the model from which Neon was adapted. Google lists Carbon API GA on November 10 and Neon API GA on November 18, 2026. Both have a unified 1,048,576-token context window.
03 // TIER PRICING
API rates by service tier
USD per 1M tokens from vendor pricing pages. 'Not published' means exactly that — we don't extrapolate.
| Service tier | Gemini 4 Neon | Gemini 4 Carbon |
|---|---|---|
| Standarddefault on-demand serving | in $3 · cached $0.15 · out $15 implicit cache only (no write fee) | Introductory through 2026-11-30 in $2 · cached $0.10 · out $10 implicit cache only (no write fee) From 2026-12-01 in $4 · cached $0.20 · out $20 implicit cache only (no write fee) |
| Batchasync jobs, 50% off | in $1.5 · cached $0.075 · out $7.5 implicit cache only (no write fee) | Introductory through 2026-11-30 in $1 · cached $0.05 · out $5 implicit cache only (no write fee) From 2026-12-01 in $2 · cached $0.10 · out $10 implicit cache only (no write fee) |
| Flexslower/cheaper serving, 50% off | in $1.5 · cached $0.075 · out $7.5 implicit cache only (no write fee) | Introductory through 2026-11-30 in $1 · cached $0.05 · out $5 implicit cache only (no write fee) From 2026-12-01 in $2 · cached $0.10 · out $10 implicit cache only (no write fee) |
| Long-context band | Gemini 4 Neon | Gemini 4 Carbon |
|---|---|---|
| Long-context ratesapplies per request above the model threshold | Flat pricing same rates at any input size (up to context window) | Flat pricing same rates at any input size (up to context window) |
USD per 1M tokens. Every input token is billed in exactly one bucket — uncached, cached read, or cache write; a write replaces the input rate for those tokens (official rule), so write rates are per tier as listed. These Gemini models use implicit caching. Enter cache hits as cached input and set explicit cache writes to zero. Gemini 4 Neon has a fixed published rate for each tier. · Google Gemini pricing (verified 2026-11-19)
Gemini service tiers
04 // COST SIMULATOR
Run this comparison on your own numbers
Preloaded with a workload typical for this comparison. Change anything — results recompute locally.
Applies to Gemini 4 Argon and Carbon. The default follows the UTC date. Google Gemini API pricing
Estimated monthly ledger
browser-localToken-based workload estimate only; it does not measure quality or cost per successful task.
Argon / Carbon rates · Introductory through 2026-11-30
| Model | Per request | Monthly |
|---|---|---|
| Gemini 4 Neon | $0.0621 | $2,484.00 |
| Gemini 4 Carboncheapest | $0.0414 | $1,656.00 |
Monthly spread between the cheapest and priciest selected model: $828.00.
Breakdown: Gemini 4 Neon · standard tier · standard context
- Fresh input (uncached): 8,000 tok × $3/M$0.024
- Cached input (reads): 4,000 tok × $0.15/M$0.0006
- Cache writes: 0 tok · rate unavailable$0.00
- Output (incl. reasoning tokens): 2,500 tok × $15/M$0.0375
- Per request$0.0621
Rates: official pricing · verified 2026-11-19
- Per-token arithmetic only: it does not model quality differences or cross-vendor output-token-count differences for the same task.
- Reasoning tokens are billed as output tokens at the chosen model's output rate.
- Long-context band is decided per request by total input tokens vs the provider threshold, not by monthly volume.
- Cache accounting uses the official mutually-exclusive buckets: each input token is billed once, as uncached input, cached read, OR cache write — writes replace the input rate for those tokens, never add to it.
- Claude rows bill the 5-minute-TTL cache write; the published 1-hour TTL ($8/MTok standard) is not modeled (no TTL input).
- Rows flagged 'derived' are computed from a stated official rule (e.g. Anthropic batch -50%), not read verbatim from a price table.
- Model and tier combinations without a published rate appear as Unsupported.
05 // BENCHMARK LOG
Benchmark results and conditions
- AutomationBench 1.0.6Gemini 4 Neon
- 54.3% (357 of 657) · high effort
Zapier API harness, 657 tasks, 50-step limit. Zapier ran Neon on November 17–18. All attempts count toward the pass rate.
- AutomationBench 1.0.6Gemini 4 Carbon
- 53.9% (354 of 657) · high effort
Zapier API harness, 657 tasks, 50-step limit. Zapier ran Carbon on October 26–27 and reused that result in the Neon report.
- AutomationBench cost per attempted task 1.0.6Gemini 4 Neon
- $0.88 per attempt · list-price estimate · high effort
Zapier priced the recorded tokens at list rates, including reasoning and cached reads. The prerelease report assumed $0.15 per million cache-hit tokens.
- AutomationBench cost per attempted task 1.0.6Gemini 4 Carbon
- $1.94 per attempt · list-price estimate · high effort
Zapier repriced Carbon usage in the Neon report’s comparison export. This export uses a different cache-read basis from the original Carbon leaderboard figure.
06 // BY WORKLOAD
Which model for which workload
Confidence labels: Measured = same-harness evidence · Price-derived = list-price arithmetic · Inferred = reasoned from docs · Insufficient evidence = no comparable public data.
| Workload | Our pick | Confidence | Why |
|---|---|---|---|
| Agent workflows | Gemini 4 Neon | Measured | Start with Neon at high effort for workflows resembling Zapier’s. It completed three more tasks at the same effort setting in the 657-task benchmark. Record attempts, successes and spend separately in your own runs. |
| Budget-sensitive / high volume | Gemini 4 Carbon | Price-derived | Through November 30, 2026, Carbon has the lower unit price. In a tool loop the models can consume different amounts of input and output. Zapier’s cost estimates show how that can reverse the ordering. |
| Complex repository maintenance | Decide by bake-off | Inferred | Google positions Carbon for sustained engineering work. Zapier graded simulated business workflows. Run your repository tasks with identical tools and stopping rules before moving them. |
QUESTIONS
Frequently asked questions
Why is Carbon cheaper in November’s calculator example but dearer in Zapier’s report?
The calculator gives both models the same token counts at the selected billing date. Zapier priced the tokens each agent actually used during its attempts. Its estimates depend on the report’s cache assumptions and do not establish production savings.
Is $0.88 the cost of a successful automation?
It is the average estimated cost across all 657 Neon attempts at high effort, including failures. Zapier reports $1.62 in estimated spend per pass when it divides the cost of all attempts by the 357 successes.
Why does Zapier also show $2.12 for Carbon?
The earlier Carbon report assumed $0.40 per million cached tokens and lists $2.12 at high effort. The Neon report quotes $1.94 for Carbon from its comparison export and flags the different cache-read basis. Both amounts are estimates.
Is Neon’s 56.3% score the high-effort result?
Zapier recorded 56.3% at xhigh. For a high-to-high comparison use 54.3% for Neon and 53.9% for Carbon. Carbon’s highest available thinking level is high.
CONTINUE EXPLORING
Related
VERIFICATION LOG
Sources & freshness
- Googleaccessed 2026-11-19Gemini 4 Neon API model documentationhttps://aistudio.google.com/docs/models/gemini-4-neonModel ID, token limits, implicit caching, thinking levels and supported tools.
- Googleaccessed 2026-11-19Gemini 4 Carbon API model documentationhttps://aistudio.google.com/docs/models/gemini-4-carbonModel ID, unified context window, maximum output, thinking levels and supported tools.
- Googleaccessed 2026-11-19Gemini API release noteshttps://aistudio.google.com/docs/changelogThe API GA entries are dated October 13 for Argon, November 10 for Carbon, November 12 for Hydrogen and November 18 for Neon, all in 2026.
- Google DeepMindaccessed 2026-11-19Gemini 4 Neon model cardhttps://deepmind.google/models/model-cards/gemini-4-neon/ · published 2026-11-18DeepMind identifies Carbon as the basis for Neon and dates the knowledge cutoff to March 2026. The evaluation table includes CursorBench 4.1 and cheating-adjusted Terminal-Bench 4.1.
- Googleaccessed 2026-11-19What's new in Gemini 4 Neonhttps://aistudio.google.com/docs/whats-new-gemini-4-neonThe Gemini 4 migration guide covers request compatibility, thinking defaults and implicit cache eligibility.
- Zapieraccessed 2026-11-19Gemini 4 Neon on AutomationBench 1.0.6https://zapier.com/benchmarks/gemini-neon-4-report · published 2026-11-18The release report includes measured pass rates and list-price estimates per attempted task. Its Carbon cost uses the report export's caching assumptions.
- Zapieraccessed 2026-11-19Gemini 4 Carbon on AutomationBench 1.0.6https://zapier.com/benchmarks/gemini-carbon-4-reportZapier tested Carbon on October 26–27, 2026. This earlier report assumed a $0.40 cache-read rate per million tokens.
- Googleaccessed 2026-11-19Gemini Developer API pricinghttps://aistudio.google.com/docs/pricingThe paid tables distinguish fresh input, implicit cache hits and output including thinking. Argon and Carbon have introductory Standard, Batch and Flex rates through November 30, 2026, with the next prices effective December 1. Google lists Priority separately.
Pricing data last verified 2026-11-19. Vendors ship frequently — re-verify before making spend commitments. See the methodology page for update cadence and stale-data flags.