FRONTIER REASONINGGPT-5.6 FAMILY1.05M CONTEXTTEXT + VISIONPROMPT CACHING
GPT-5.6 Sol API Pricing
GPT-5.6 Sol is OpenAI's highest-accuracy GPT-5.6 tier for hard reasoning, coding and senior-agent workloads, and it has now had a cut of its own: $4/M input, $20/M output and $0.40/M cached input, against $5 / $30 / $0.50 before. Input-side rates fell to 0.8x and output to two thirds, taking the effective agentic rate from $3.81 to $2.73/M. OpenAI labels the new card promotional and now says the promotion ends on November 30, 2026, without publishing what comes after. Sol sat out the July 30 cut that took Luna down 80% and Terra 20% - this one is its own, and it is the only GPT-5 row that moved: Terra, Luna, 5.5, 5.4 and both -pro tiers read unchanged on the same page.
Input - per 1M tokens
$4.00/M
Short context up to 272K input standard
Output - per 1M tokens
$20.00/M
Long context output is $30/M standard
Cached input - 90% off
$0.40/M
Prompt cache read price -90%
Effective - agentic blend
$2.73/M
92/8 split - 82% cache
§ 01 / TERMINAL
GPT-5.6 Sol API cost calculator.
Live calculator pre-loaded with current GPT-5.6 Sol standard short-context rates. Tweak spend, output mix, or cache hit rate; share the URL to share the calculation.
$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS
GPT-5.6 Sol cost on real workloads.
CODING AGENT
Repo-wide architecture migration
$0.387/task
RESEARCH
Multi-step diligence brief
$0.154/brief
LONG DOC ANALYSIS
Reading 150-page contracts
$0.455/doc
ASSISTANT
Executive workflow turn
$0.012/turn
Exact · o200k_base Auto-counts as you type
Counted with the o200k_base BPE encoding — the same tokenizer the API uses — entirely in your browser. The encoder loads on your first keystroke (~1 MB, one time); your text never leaves this page.
Characters 577
Words 86
Tokens (exact) 151 tokens
Cost as input · uncached $0.0006 USD
Cost as output · uncached $0.00302 USD
Cost as cached input $0.00006 USD
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| GPT-5.6 Sol Current | $4.00 cache $0.40 | $20.00 | $2.73 agentic 92/8 | 1.05M | Highest-accuracy GPT-5.6 preview |
| GPT-5.6 Terra | $2.00 cache $0.20 | $12.00 | $1.52 −44% | 1.05M | Balanced GPT-5.6 preview agents |
| GPT-5.6 Luna | $0.20 cache $0.02 | $1.20 | $0.152 −18× | 1.05M | Cheapest GPT-5.6 preview routing |
| GPT-5.5 | $5.00 cache $0.50 | $30.00 | $3.61 +32% | 1M | Current broadly available OpenAI flagship |
| GPT-5.4 | $2.50 cache $0.25 | $15.00 | $1.80 −34% | 1.05M | Broadly available balanced OpenAI tier |
| GPT-5.4 mini | $0.75 cache $0.075 | $4.50 | $0.541 −5.0× | 400K | Lower-cost OpenAI subagents |
| Claude Opus 4.8 | $5.00 cache $0.50 | $25.00 | $3.41 +25% | 1M | Anthropic Opus frontier alternative |
| Gemini 3.6 Flash | $0.75 cache $0.075 | $3.75 | $0.481 −5.7× · promo | 1M | Google multimodal context alternative |
Compare siblings
GPT-5.6 Sol has published scores on 15 benchmarks we track; its best placing is #4 of 105 on ARC-AGI-2. 4 of these scores come from an independent leaderboard; the rest are the lab's own reports.
| Benchmark | Score | Rank | Leader | Source |
|---|---|---|---|---|
| ARC-AGI-2 Reasoning | 92.5% ARC Prize verified, best result | #4 of 105 | GPT-6 Astra · 95% | Official leaderboard ARC Prize Foundation |
| ARC-AGI-1 Reasoning | 97.5% ARC Prize verified, best result | #8 of 99 | 6 tied · 98.5% | Official leaderboard ARC Prize Foundation |
| GPQA Diamond Expert knowledge | 94.6% as reported | #2 of 11 | GPT-6 Astra · 96% | Vendor-reported OpenAI |
| ExploitBench Cybersecurity | 78.5% without production safeguards | #2 of 7 | GPT-6 Astra · 100% | Vendor-reported OpenAI |
| ARC-AGI-3 Reasoning · Standard harness | 7.8% ARC Prize verified, best result | #7 of 24 | GPT-6 Astra · 62.7% | Official leaderboard ARC Prize Foundation |
| DeepSWE Coding | 72.7% max effort | #8 of 27 | Gemini 4 Argon · 77.9% | Vendor-reported OpenAI |
| ExploitGym Cybersecurity | 30.3% without production safeguards | #3 of 9 | Gemini 4 Carbon · 54.8% | Vendor-reported OpenAI |
| SEC-bench Pro Cybersecurity | 79.1% as reported | #3 of 8 | Gemini 4 Carbon · 88.2% | Vendor-reported OpenAI |
| Terminal-Bench 4.0 Agents and computer use | 37.3% max effort, Codex harness, 330 trials, ±3.8 pts (95% CI) | #17 of 34 | Kimi K3.1 Pro · 65.7% | Official leaderboard Terminal-Bench (Stanford, Laude Institute) |
| FrontierCode Coding · v1.1 Main (100 hardest tasks) | 47.5% max effort | #7 of 13 | Claude Opus 5.5 · 54.4% | Vendor-reported OpenAI |
| OSWorld 2 Agents and computer use · 2.0, offline subset, partial score | 66.2% max effort | #4 of 7 | GPT-6 Astra · 73.5% | Vendor-reported OpenAI |
| Agents' Last Exam Agents and computer use · ALE v1 | 53.6% xhigh effort | #3 of 5 | GPT-6 Astra · 59.3% | Vendor-reported OpenAI |
| BrowseComp Agents and computer use | 90.4% as reported | #4 of 6 | Kimi K3.1 Pro · 92.1% | Vendor-reported OpenAI |
| AutomationBench Agents and computer use · Main set (as reported) | 28.8% max effort, v1.0.6 | #16 of 22 | MiniMax M3.1 Flash · 62.8% | Vendor-reported OpenAI |
| Terminal-Bench-Science AI research | 22.4% as reported | #12 of 12 | Claude Mythos 5.2 · 75.3% | Vendor-reported OpenAI |
GPT-5.6 Sol API pricing FAQ.
Practical GPT-5.6 Sol pricing questions, with OpenAI's published rates separated from workload assumptions.
Q · 01 How much does GPT-5.6 Sol cost? +
OpenAI lists GPT-5.6 Sol at $4/M input tokens, $0.40/M cached input tokens, $5/M cache writes and $20/M output tokens in the standard short-context tier. That is a cut from $5 / $0.50 / $6.25 / $30, and the page marks it promotional until November 30, 2026. Verified on OpenAI's pricing docs 2026-11-07.
Q · 02 Is GPT-5.6 Sol generally available? +
Yes. OpenAI made the GPT-5.6 series (Sol, Terra, Luna) generally available on July 9, 2026 across ChatGPT, Codex, and the API, after an initial limited preview that opened June 26, 2026.
Q · 03 What is the context window? +
OpenAI's model docs list a
1.05M-token context window and 128K max output for GPT-5.6 Sol, with a stated knowledge cutoff of February 16, 2026.Q · 04 How much cheaper are Batch and Flex? +
OpenAI lists Batch and Flex for GPT-5.6 Sol at half the standard short-context rate:
$2/M input, $0.20/M cached input, $2.50/M cache writes and $10/M output. Fast mode - the tier renamed from Priority on July 30, 2026 - is double standard: $8/M input, $0.80/M cached input, $10/M cache writes and $40/M output.Q · 05 When does long-context pricing apply? +
For requests above
272K input tokens, GPT-5.6 Sol uses the long-context tier: $8/M input, $0.80/M cached input, $10/M cache writes and $30/M output — double the standard input rate and 1.5x the standard output rate, the same multipliers as before the cut. The calculator on this page uses the standard short-context tier.Q · 06 Which GPT-5.6 tier should I choose? +
Use Sol for the hardest tasks, Terra for balanced agent workloads, and Luna for lower-cost routing or high-volume production. In the shared 92/8 agentic blend with 82% cache hits, Sol now lands near
$2.73/M, Terra near $1.52/M, and Luna near $0.15/M. The August cut narrowed the family: Luna is 17.9x cheaper than Sol, down from 25x, and Sol is only 1.8x Terra rather than 2.5x - which is exactly the sort of change that makes an old routing rule wrong.Q · 07 Did Sol get cheaper in the 2026 price cuts? +
Not in July, but yes in August. The July 30, 2026 cut took Luna down 80% and Terra 20% and left Sol at its launch rate, widening the gap to 25x. On or before August 23 Sol had its own cut - to
$4 / $0.40 / $5 / $20 - which brings the blended agentic rate to $2.73/M and narrows the gap to Luna to 17.9x. Unlike July, this one comes labelled promotional, which OpenAI now says ends on November 30, 2026, with no published rate for what follows.