LeadsCalc
Interactive leaderboard

Best Long-Context LLMs 2026: Large-Window AI APIs

Explore long-context LLMs in 2026: 1M+ token context windows, RAG-friendly SKUs, and estimated API pricing. Built for legal, docs, and enterprise RAG in the US, Canada, and Australia.

Long-context LLMs ranked for RAG, docs, and codebases in 2026

Large context reduces chunking pain for legal bundles, multi-file repos, and executive briefs—but token cost scales with what you paste. This view foregrounds window size and fitness for retrieval-heavy stacks while keeping monthly estimates honest for teams in the US, Canada, and Australia planning enterprise rollouts.

Est. monthlyROI scoreCodingReasoningSpeedMathContextOverallOpen-weight

Workload & pricing toggles

Workload presets

Same three scenarios as the main AI API calculator: moderate traffic, large RAG-style context, or per-request max tokens with a lower request count.

Include Vision / Image Processing

Off — no image fees in cost estimates for vision-capable models.

Turn On to include image fees.

OffOn

Use Cached Pricing

Enable to get 50% off input tokens where cached rates apply

OffOn

Deep Reasoning / Thinking Mode

Model hidden reasoning / extended thinking charged like output tokens when enabled.

OffOn

Batch Pricing

Enable for 50% off input & output where batch/async pricing applies

OffOn
≈ $100.00/mo
8K
1K—1.0M
≈ $100.00/mo
2K
100—500K
≈ $200.00 total
5K
10—100K

Cached / batch est. monthly values only change after the pipeline sets supports_caching or supports_batch in Supabase. The toggles here narrow the table to models whose catalog or provider typically supports those modes.

Magic quadrant (top 15)

X: est. monthly · Y: Long context · Dot: provider color · Hover for rank, model & details

Full leaderboard

Showing 48 of 513 models.

PickModelEst. monthlyROI scoreCodingReasoningSpeedMathContextOverall
xAI: Grok 4.1 Fast$13.0060
75
83
90
75
2.0M
79Evidence lacks raw benchmarks. Inferred scores based on 'Fast' lightweight tier and agentic focus. Speed rated high (90); coding (75) and logic (80) adjusted lower than flagship models. Native reasoning enabled.
xAI: Grok 4 Fast$13.0040
40
43
95
45
2.0M
43Evidence cites a 43.5 average across HumanEval, GPQA, and MMLU for this 1B-scale Fast model. Mapped to ~40-45 for coding and logic. As a lightweight tier, speed is rated very high.
Auto Router (Beta)
VARIABLE
77
85
85
70
85
2.0M
85Primary source: OpenRouter docs (no raw scores). Inferred frontier-level capabilities (~85) across coding, logic, instruction, and math based on task-aware routing to popular models. Speed estimated at 70. Reasoning explicitly omitted for dynamic routers.
Auto Router
VARIABLE
80
90
90
70
90
2.0M
90Auto Router optimizes across models for best output. Evidence cites top models reaching 92.3% MMLU. Mapped to 90 across logic, coding, and math to reflect frontier routing capabilities. Vision price defaulted to $0.007 per tier guidelines.
SpaceXAI: Grok 4.20 Multi-Agent$75.0058
85
87
45
85
2.0M
86Evidence lacks explicit Grok benchmarks (listed as 'Not available'). Inferred as a 2026 flagship reasoning model competing with Opus 4.6. Scores estimated for a heavy multi-agent reasoning tier; speed is lower due to 16-agent coordination.
SpaceXAI: Grok 4.20$75.0061
95
91
80
90
2.0M
92SWE-bench (78.0%) maps to 95 coding. GPQA Diamond (74.5%) maps to 92 logic. As a flagship model, instruction and math align with frontier scores. Speed is high (80) due to being the low-latency, non-reasoning variant.
Pareto Code Router
VARIABLE
77
88
85
70
85
2.0M
86OpenRouter docs state this is a router defaulting to High tier coding models based on Artificial Analysis percentiles. Lacking specific raw benchmarks, scores are mapped to ~85 reflecting flagship-level routed performance. Text-only inputs confirmed.
Llama 4 Scout$7.0059
65
73
85
70
1.3M
70MMMU 69.4% and ChartQA 88.8% map to 70 multimodal. Lacking SWE-bench or GPQA, coding and logic are inferred (65-70) for this 17B-active lightweight MoE tier. Speed is high (85) due to small active parameter count.
OpenAI: GPT-6.1 Sol$180.0058
75
92
75
98
1.1M
89Based on GPT-5.6 Sol evidence: SWE-bench 74.6% (Coding 75), GPQA 89.6% (Logic 90), MATH 97.8% (Math 98). Positioned below flagship Astra, mapped as a strong mid-tier model with native reasoning.
OpenAI: GPT-5.6 Luna$20.0051
65
68
90
65
1.1M
66Evidence lacks exact benchmarks for GPT-5.6 Luna. Inferred scores from its 'fast, cost-efficient' lightweight tier description, prioritizing speed (90) while scaling down coding and logic versus flagship models.
OpenAI: GPT-6 Luna Pro$9.0062
85
75
95
75
1.1M
78DeepSWE 66.6% and HumanEval 82.4% map to 85 coding. MMLU-Pro 74.6% maps to 75 logic. Speed of 157 t/s maps to 95. As a lightweight reasoning model, scores reflect its tier despite high SWE-bench performance.
OpenAI: GPT-5.4 Pro$3,000.0057
95
93
60
92
1.1M
93GPT-5.4 Pro lacks exact SWE-bench/GPQA scores but beats Gemini 3.1 Pro on SWE-Bench Pro and MMMU-Pro. Inferred as a flagship model: Coding ~95, Logic ~92 (GDPval 83%). Multimodal inferred ~85.
OpenAI: GPT-5.6 Luna Pro$20.0066
92
93
45
95
1.1M
93Evidence lacks exact GPT-5.6 Luna Pro benchmarks. Inferred scores based on its 'Pro' heavyweight tier and native reasoning mode. Assigned high logic (95) and coding (92), with lower speed (45) typical for extended reasoning models.
OpenAI: GPT-5.6 Terra Pro$200.0059
90
91
45
92
1.1M
91Web digest notes 37 tokens/s (Speed 45). Lacking exact GPT-5.6 Terra Pro benchmarks, inferred flagship 'Pro' tier with native reasoning: Coding 90, Logic 92. Vision price defaulted to $0.007.
OpenAI: GPT-6 Astra$900.0061
98
97
55
99
1.1M
98DeepSWE 74.1% maps to 98 coding. GPQA Diamond 96% maps to 99 logic. FrontierMath 97.6% maps to 99 math. MMMU 74.8% maps to 85 multimodal. Flagship reasoning model speed is estimated at 55.
OpenAI: GPT-6 Sol Pro$180.0060
92
93
40
95
1.1M
93Evidence lacks exact GPT-6 Sol Pro benchmarks. Inferred flagship reasoning tier (Pro) for high logic/coding/math (~95). Speed is low (~40) due to 'takes much longer' reasoning tokens. Multimodal inferred (~85) from native vision support.
OpenAI: GPT-6 Sol$180.0052
75
88
86
64
1.1M
79BenchLeader cites reasoning at 85 (Logic: 85) and math at 64 (Math: 64). Speed is 86 tok/s (Speed: 86). Coding (75) and Multimodal (83) inferred from predecessor GPT-5.6 Sol's SWE-bench (74.6) and MMMU (83.1).
Xiaomi: MiMo-V2.5-Pro$18.2765
95
91
65
88
1.1M
91SWE-bench Verified 78.9% maps to 95 coding. GPQA Diamond 66.7% maps to 92 logic. Flagship tier model with native reasoning and caching support.
OpenAI: GPT Astra Latest$900.0054
90
89
71
76
1.1M
86Flagship tier. GPQA Diamond 96.1% maps to logic 96. HumanEval 90.2% maps to coding 90. IFEval 81% maps to instruction 81. MATH 75.9% maps to math 76. Speed 71.3 t/s maps to 71. MMMU ~74% maps to multimodal 72.
OpenAI: GPT-6.1 Sol Pro$180.0058
75
92
40
98
1.1M
89SWE-bench at 74.6 maps to 75 coding. GPQA at 89.6% maps to 90 logic. MATH at 97.8% maps to 98 math. As a 'Pro' reasoning model, speed is mapped lower (40).
Xiaomi: MiMo-V2.5$7.1467
78
86
60
85
1.1M
84Pro-level agentic performance maps to MiMo-V2-Pro's SWE-bench Verified (78.0%) and GPQA Diamond (87.0%), yielding ~78 Coding and ~87 Logic. Speed reflects 29 tok/s. Multimodal is strong per omnimodal claims.
Xiaomi: MiMo-V2.6-Pro$26.1062
85
85
85
95
1.1M
88DeepSWE 71.9% and GPQA 66.7% (V2.5) map to 85 coding and logic. GSM8K 99.6% maps to 95 math. 120.5 tokens/s maps to 85 speed. Flagship tier justifies strong baseline scores.
Xiaomi: MiMo-V2.6-Flash$8.4061
88
70
90
75
1.1M
76SWE-bench Verified reported at 78.6% maps to 88 coding. GPQA-Diamond at 55.1% (V2 baseline) maps to 60 logic. As a Flash-tier model, speed is rated high (90), while reasoning scores reflect its lightweight, efficiency-focused architecture.
OpenAI: GPT Sol Latest$180.0062
96
97
45
99
1.1M
97SWE-bench Pro at 64.6% maps to 96 coding. GPQA Diamond at 94.6% maps to 98 logic. FrontierMath at 89.0% maps to 99 math. Speed is 45 based on 27.2 tokens/s throughput for this flagship reasoning model.
OpenAI GPT Latest$180.0055
85
86
77
80
1.1M
84Based on GPT-4.1 data: SWE-bench Verified 54.6% maps to 85 coding, GPQA 66.3% maps to 85 logic, IFEval 87.4% maps to 87 instruction. MMMU 74.8% maps to 80 multimodal. Flagship tier, no lightweight adjustment.
OpenAI: GPT-5.6 Sol$360.0058
91
88
60
97
1.1M
91Lacking explicit GPT-5.6 Sol benchmarks, inferred flagship scores from top leaderboard entries (SWE-bench ~91%, GSM8K ~97%) and GPT-4.5 GPQA (71.4%). Mapped to 85-97 range for logic/coding/math. Speed set to 60 for heavyweight tier.
OpenAI: GPT Terra Latest$200.0058
88
90
80
90
1.1M
90SWE-bench Pro 63.4% maps to 88 coding. GPQA Diamond 92.9% maps to 95 logic. MMMU-Pro 80.7% maps to 88 multimodal. As a balanced mid-tier model, speed is 80. Native reasoning and caching are supported.
OpenAI: GPT-6 Luna$9.0061
85
75
95
70
1.1M
76DeepSWE 1.1 at 66.6% and HumanEval at 82.4% map to 85 coding. MMLU-Pro at 74.6% maps to 75 logic. As a lightweight 'Luna' tier, scores reflect its cost-efficiency focus despite high reasoning effort.
OpenAI: GPT-5.4$250.0059
95
91
70
88
1.1M
91Evidence states GPT-5.4 outperforms GPT-4.1 (SWE-bench Verified 54.6%, GPQA 66.3%). Mapped coding to 95 and logic to 92 for this frontier flagship. Multimodal inferred from native computer-use screenshots and GPT-4.1's MMMU 74.8%.
OpenAI: GPT-5.5$500.0056
85
89
60
90
1.1M
88Evidence states GPT-5.5 outperforms GPT-4.1 (SWE-bench Verified 54.6%, GPQA 66.3%). As a frontier model with native real-time reasoning, scores are mapped to elite flagship tiers (85-90+). Speed is typical for heavyweights.
OpenAI: GPT Luna Latest$9.0064
82
81
95
85
1.1M
82SWE-bench Pro at 62.7% maps to 82 coding. GPQA at 92.3% maps to 92 logic. As a lightweight Luna tier model, speed is rated 95, balancing high-volume latency with strong reasoning capabilities.
OpenAI: GPT-5.5 Pro$3,000.0057
90
94
60
92
1.1M
92LLM Benchmarks reports 94.8 overall score, mapped to 95 logic. Outperforms in GPQA and MathVista. As a 1T parameter Pro flagship, coding and math are estimated at 90-92. Speed is standard for heavyweights (60).
OpenAI: GPT-6 Astra Pro$900.0062
98
99
45
99
1.1M
99DeepSWE 74.1% maps to 98 coding. GPQA Diamond 96% maps to 99 logic. FrontierMath 97.6% maps to 99 math. MMMU 74.8% maps to 85 multimodal. Flagship reasoning model with native effort controls, yielding lower speed (45).
OpenAI: GPT-5.6 Terra$200.0053
80
81
80
80
1.1M
81Specific GPT-5.6 Terra metrics are proprietary. Inferred scores based on its balanced mid-tier positioning between Sol and Luna. Assigned ~80s across coding, logic, and math, with speed at 80 and multimodal at 75.
OpenAI: GPT-5.6 Sol Pro$360.0059
92
93
45
94
1.1M
93No exact benchmarks for GPT-5.6 Sol Pro in evidence. Inferred flagship scores (90+) based on 'Pro' tier and 'reasoning.mode' capabilities. Speed adjusted lower (~45) for dedicated reasoning mode.
Owl Alpha
Free
66
65
68
85
60
1.0M
65No exact scores for Owl Alpha; inferred as a lightweight reasoning model ('fewer parameters', 'designed for speed'). Mapped to mid-tier 0-100 scale (Coding 65, Logic 65) reflecting its agentic focus but smaller size.
Meituan: LongCat 2.0$24.0063
85
85
65
95
1.0M
88HumanEval 78.5% and GSM8K 92% cited for Flash Chat; as the 1.6T flagship, LongCat 2.0 maps higher (Coding 85, Math 95). Logic and Instruction estimated at 85. Speed 65 for heavy MoE.
Ox Alpha
Free
79
90
88
85
85
1.0M
88Primary source: Web digest (102.5 tok/s). Lacking exact SWE-bench or GPQA percentages for Ox Alpha, capabilities are inferred for a flagship reasoning tier. Speed mapped to 85 based on the cited high throughput.
Mistral: Mistral Large 4$96.2054
85
80
65
76
1.0M
80Lacking Large 4 benchmarks, inferred from Large 2 (HumanEval 92%, GPQA ~25%, MMLU 84%, GSM8K 76%). Mapped Coding to 85, Logic to 80, Math to 76. Flagship tier speed (65) and multimodal (82).
MoonshotAI: Kimi K3.1 Pro$180.0056
85
85
40
85
1.0M
85Inferact README cites SWE-bench Pro 3.35 and MT-Bench 3.14. Mapped to 85 coding and 80 instruction. Bybit Wiki confirms o3/R1-class CoT reasoning, justifying high logic (90) but lower speed (40) for this flagship model.
Z.ai: GLM 5.3$17.1662
78
87
45
85
1.0M
84GLM-5 series scores SWE-bench Verified 77.8% (Coding 78) and GPQA Diamond 86.0% (Logic 86). IFEval is 88.0% (Instruction 88). MMMU inclusion confirms multimodal. As a 744B reasoning flagship, speed is lower (~45).
Pareto 26.10 Preview$64.0057
70
89
60
85
1.0M
83GPQA Diamond at 92.4% maps to 92 logic. Terminal-Bench 4.0 at 50.80% maps to 70 coding. As a composite frontier model, speed is estimated at 60. Multimodal inferred from tier.
Google: Gemini 3.5 Flash Lite$37.0053
70
73
95
75
1.0M
73Lacking exact 3.5 Flash-Lite scores, inferred from Gemini 3 Flash (SWE-bench Verified 78.0%, GPQA 90.4%). Mapped to 0-100 scale with strict penalties for the 'Lite' tier, reducing coding and logic while maximizing speed.
Google: Gemini Flash Latest$67.5049
70
65
95
75
1.0M
69Evidence cites HumanEval 74.3% (Coding ~70), GPQA 51.0% (Logic ~60), and MMMU 62.3% (Multimodal ~65). As a lightweight Flash tier, scores are adjusted lower than Pro flagships, while Speed is rated high (~95) for its class.
Fireworks: Ember-1$270.0058
92
90
45
87
1.0M
90SWE-bench Verified at 92.2% maps to 92 for coding. GPQA proxy signal of 95 maps to 95 for logic. As a dedicated reasoning model, speed is rated lower (45). No vision capabilities or caching/batch features were documented.
Google: Gemini 4 Argon$180.0052
75
75
45
85
1.0M
78Based on GPQA (51.0%) for logic (70) and HumanEval (74.3%) for coding (75). Math mapped to 85 from MATH (77.9%). As a flagship reasoning model, speed is lower (45). Multimodal scored 65 via MMMU (62.3%).
Google: Gemini 2.5 Flash Lite Preview 09-2025$8.0052
45
67
95
55
1.0M
58Based on GPQA (65.1-70.9%) and LiveCodeBench (64.1-68.8%), logic and coding map to 68 and 45. As a 'Flash Lite' tier, speed is heavily weighted (95), reflecting its ultra-low latency design over flagship-level reasoning.
Google: Gemini 3.1 Pro Preview Custom Tools$200.0062
98
97
65
95
1.0M
97SWE-bench Verified at 80.6% maps to 98 coding. GPQA Diamond at 94.3% maps to 98 logic. As a flagship Pro model, it receives high multimodal (95) and math (95) scores, with standard heavyweight speed (65).

Need a shareable artifact?

Get a print-ready PDF of your results and a CSV spreadsheet. Tap the button, then enter your work email. We use it to build your files and start the download—and to email you a copy if the site owner enabled that.

AI ROI Leaderboard & Discovery by LeadsCalc

Detailed analysis

PDF Breakdown

Receive a comprehensive native vector PDF of this leaderboard: your workload, filters, top rankings, and a table snapshot (sorted: Long context).

Instant setup
No CC required

By submitting, you agree to our Privacy Policy and Terms.

Agency accelerator

Whitelabel Context Leaderboard
for your site

Embed the interactive long context view on your own domain — whitelabel branding, lead capture, and the same workload sliders your prospects already use on LeadsCalc.

1-Click CRM sync
Custom branding
Branded reports
Lead analytics

Free to start

$0/mo*
GET STARTED

NO CREDIT CARD REQUIRED

How it works

Methodology: How we rank Long-Context LLMs

Transparent, benchmark-driven rankings—same craft as our single-model deep dives.

Context window data and how it interacts with price

Our long-context rankings evaluate models based on their maximum supported context window (ranging from 128k to 2M+ tokens) and their proven ability to retrieve information accurately at high fill rates (Needle In A Haystack performance). This view is purpose-built for teams building enterprise RAG systems, legal document analyzers, and repository-scale coding assistants.

Battle Arena

Compare up to four LLMs side by side

Tick up to four models in the leaderboard table, then open Battle Arena for API pricing, benchmarks, and workload math in one view—perfect when you are shortlisting vendors for a pilot in the US, Canada, or Australia.

Prefer a head start? Jump into high-intent comparisons people search for every day—same interactive calculator, zero signup.

Open Battle ArenaUp to 4 models · Live estimates
Signals & spend

Value analysis

Benchmarks vs. estimated API cost—read the story your CFO cares about.

When a mega-context model beats clever chunking

If your prompts repeat the same long system preamble, prompt caching may matter as much as raw window size—toggle cached pricing when supported. Canadian and Australian enterprises often evaluate sovereignty and subprocessors alongside context; US teams may prioritize vendor BAAs and retention policies.

Production deployment

Retrieval-Augmented Generation (RAG)

How teams in the US, Canada, and Australia deploy these models in production.

Needle-in-a-haystack Q&A and book-length summarization

Models with 1M+ token windows are transforming how enterprises handle unstructured data. Legal teams in the US and compliance officers in Canada use these massive-context models to ingest entire case files, financial prospectuses, or sprawling codebases in a single prompt, bypassing the complexity and retrieval loss associated with traditional vector database chunking.

Architecture

Long-Context Cost Management

Strategies to reduce monthly API spend without sacrificing capability.

Prefix caching economics and context compression

Sending 500,000 tokens per request is financially ruinous without optimization. The key to affordable long-context architecture is Prompt Caching. By keeping the massive document at the start of the prompt (the prefix), providers can cache it and discount subsequent queries by up to 90%. Ensure you toggle 'Use Cached Pricing' on this leaderboard to accurately forecast long-context RAG costs.

Embed-ready

Need this live Long-Context data on your website?

Join 500+ agencies in the US and Australia using LeadsCalc to capture high-intent leads. Embed this interactive Long-Context leaderboard on your site in about a minute—Canadian teams use the same flows for CAD-priced proposals and compliance-friendly landing pages.

Customize & Embed this ToolWhite-label · No code required
United StatesCanadaAustralia
Live preview

Your visitors compare Long-Context models without leaving your domain.

Support & clarity

Frequently Asked Questions

Focused on teams in the United States, Canada, and Australia.

Models with 1M+ token windows (like Gemini 1.5 Pro or Claude 3.5 Sonnet) are currently the industry standard for massive document analysis. They allow teams in the US, Canada, and Australia to ingest entire legal briefs or financial prospectuses in a single prompt, bypassing the complexity of traditional vector database chunking.