GPT-5.2 vs Sarvam 30B

Updated October 17, 2026. Rank says GPT-5.2 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Decision reading

GPT-5.2 has the higher public point estimate, 59.86 versus 28.61. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 2 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A

GPT-5.2

OpenAI

59.86/100

Supported · Public rank #56

90% interval 52.7–67.0

Model B

Sarvam 30B

Sarvam

28.61/100

Estimated · Public rank #186

Conditional range 14.3–43.0

Shared results: 2
GPT-5.2 only: 29
Sarvam 30B only: 7
Like-for-like categories: 1 / 8

Supported: GPT-5.2 · Estimated: Sarvam 30B. Conditional ranges do not establish rank confidence.How the comparison works

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

Long documents
Prompts that approach the documented context limit
GPT-5.2
GPT-5.2 has the larger documented context window.
Confidence: documented

Show secondary and unsupported calls

Coding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
Sarvam 30B is not ranked on the public lane for coding, so no winner is named for coding.
Confidence: limited
Agentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
Sarvam 30B is not ranked on the public lane for agentic, so no winner is named for agentic.
Confidence: limited
Chat turn cost
1K fresh input + 500 output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates
Cache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Sarvam 30B does not fit this workload in one request. GPT-5.2 has no published cached-input rate, so cached tokens use its listed input rate. Sarvam 30B has no comparable published API token rate.
Confidence: rate-fallback
Repository review cost
50K fresh input + 3K output tokens
Not enough matched evidence
A complete comparable API-rate estimate is not available for both models.
Confidence: listed-rates

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

Pick the job you would use them for

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

36.8GPT-5.2—Sarvam 30B

Not comparable · BenchAlign v5.8

The coding row is not comparable on the public lane: at least one model is not measured or not ranked there.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboard Shortlist models for this job

Same basis rules as the category table below

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

SWE-bench VerifiedCoding
Normalized gap 46.0
GPT-5.2:80%
Sarvam 30B:34%
GPT-5.2 source Sarvam 30B source
BrowseCompAgentic
Normalized gap 30.3
GPT-5.2:65.8%
Sarvam 30B:35.5%
GPT-5.2 source Sarvam 30B source

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Knowledge

Like-for-like

GPT-5.2: 59.7; Supported · #50/184
Sarvam 30B: 27.2; Supported · #165/184
Basis: BenchAlign v5.8 lane · 1 vs 2 public rows
Reading: GPT-5.2 leads

Instruction following

Directional only

GPT-5.2: 92.6; #19/148
Sarvam 30B: 28.8; #140/148
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Directional only

Agentic

Not comparable

GPT-5.2: 44.4; Supported · #58/128
Sarvam 30B: Not ranked
Basis: BenchAlign v5.8 lane · 4 vs 1 public rows
Reading: Not comparable

Coding

Not comparable

GPT-5.2: 36.8; Supported · #73/150
Sarvam 30B: Not ranked
Basis: BenchAlign v5.8 lane · 3 vs 2 public rows
Reading: Not comparable

Reasoning

Not comparable

GPT-5.2: 64.3; Unranked · 4 rankable rows
Sarvam 30B: 25.9; Unranked · 3 rankable rows
Basis: Provisional lane · 1 vs 0 weighted rows
Reading: Not comparable

Multimodal

Not comparable

GPT-5.2: 75.9; #22/54
Sarvam 30B: Not ranked
Basis: Provisional lane · 2 vs 0 weighted rows
Reading: Not comparable

Multilingual

Not comparable

GPT-5.2: Not ranked
Sarvam 30B: Not ranked
Basis: Provisional lane · 0 vs 0 weighted rows
Reading: Not comparable

Math

Not comparable

GPT-5.2: 71.1; Unranked · 3 rankable rows
Sarvam 30B: Not ranked
Basis: Provisional lane · 2 vs 0 weighted rows
Reading: Not comparable

Category scores with the server-provided evidence basis for GPT-5.2 and Sarvam 30B
Category	GPT-5.2	Sarvam 30B	Basis	Reading
Knowledge	59.7Supported · #50/184	27.2Supported · #165/184	Like-for-likeBenchAlign v5.8 lane · 1 vs 2 public rows	GPT-5.2 leads
Instruction following	92.6#19/148	28.8#140/148	Directional onlyProvisional lane · 0 vs 0 weighted rows	Directional only
Agentic	44.4Supported · #58/128	Not ranked	Not comparableBenchAlign v5.8 lane · 4 vs 1 public rows	Not comparable
Coding	36.8Supported · #73/150	Not ranked	Not comparableBenchAlign v5.8 lane · 3 vs 2 public rows	Not comparable
Reasoning	64.3Unranked · 4 rankable rows	25.9Unranked · 3 rankable rows	Not comparableProvisional lane · 1 vs 0 weighted rows	Not comparable
Multimodal	75.9#22/54	Not ranked	Not comparableProvisional lane · 2 vs 0 weighted rows	Not comparable
Multilingual	Not ranked	Not ranked	Not comparableProvisional lane · 0 vs 0 weighted rows	Not comparable
Math	71.1Unranked · 3 rankable rows	Not ranked	Not comparableProvisional lane · 2 vs 0 weighted rows	Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

GPT-5.2: $0.00875; Fits in one request
Sarvam 30B: Self-hosted; infrastructure cost varies; Fits in one request

Sarvam 30B has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

GPT-5.2: $0.1295; Fits in one request
Sarvam 30B: Self-hosted; infrastructure cost varies; Fits in one request

Sarvam 30B has no comparable published API token rate.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

GPT-5.2: $0.525; Fits in one request; Cached input priced at the published list-input rate
Sarvam 30B: Self-hosted; infrastructure cost varies; Does not fit in one request; Cached-input rate unavailable

Sarvam 30B does not fit this workload in one request. GPT-5.2 has no published cached-input rate, so cached tokens use its listed input rate. Sarvam 30B has no comparable published API token rate.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

SpecificationGPT-5.2Sarvam 30B

Context window

Maximum documented context; output-token limits may be lower.

GPT-5.2

400K

Sarvam 30B

64K

API model ID

GPT-5.2

Not sourced

Sarvam 30B

Not sourced

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

GPT-5.2

Not published

Sarvam 30B

No comparable hosted API rate

Documented inputs

GPT-5.2

Not sourced

Sarvam 30B

Not sourced

Documented outputs

GPT-5.2

Not sourced

Sarvam 30B

Not sourced

Provider availability

GPT-5.2

Not sourced

Sarvam 30B

Not sourced

Reasoning profile

GPT-5.2

Reasoning

Sarvam 30B

Reasoning

Weight access

GPT-5.2

Proprietary

Sarvam 30B

Open Weight

License

GPT-5.2

Proprietary

Sarvam 30B

Open Weight

Release date

GPT-5.2

2025-12-11

Sarvam 30B

2026-03-06

If you already use one of these models

Deployment change: The models list different providers, so authentication, endpoint behavior, limits, and feature support may change.
Quality signal: GPT-5.2 has the higher public point estimate, 59.86 versus 28.61. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost: A complete comparable API-rate estimate is not available for both models.
Context tradeoff: GPT-5.2 has the larger documented window (400K).

Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, GPT-5.2 or Sarvam 30B?

GPT-5.2 has the higher public point estimate, 59.86 versus 28.61. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, GPT-5.2 or Sarvam 30B?

Sarvam 30B is not ranked on the public lane for coding, so no winner is named for coding.

Which is better for agentic tasks, GPT-5.2 or Sarvam 30B?

Sarvam 30B is not ranked on the public lane for agentic tasks, so no winner is named for agentic tasks.

Which costs less, GPT-5.2 or Sarvam 30B?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, GPT-5.2 or Sarvam 30B?

GPT-5.2 has the larger documented context window: 400K, compared with 64K.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

GPT-5.2

API / mo$11,813

Self-host / moNot listed

Break-even—

Proprietary model — self-hosting not applicable.

Sarvam 30B

API / mo$0

Self-host / mo$298

Break-even—

Model the full break-even

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence38 rows

Agentic

BrowseComp
GPT-5.265.8%
Source
Sarvam 30B35.5%
Source
GPT-5.2 leads this result
OSWorld-Verified
GPT-5.247.3%
Source
Sarvam 30B—
Not directly comparable
Gert Labs
GPT-5.246.54%
Source
Sarvam 30B—
Not directly comparable
JobBench
GPT-5.234.3%
Source
Sarvam 30B—
Not directly comparable

Coding

SWE-bench Verified
GPT-5.280%
Source
Sarvam 30B34%
Source
GPT-5.2 leads this result
SWE-bench Pro
GPT-5.255.6%
Source
Sarvam 30B—
Not directly comparable
Vibe Code Bench
GPT-5.253.50%
Source
Sarvam 30B—
Not directly comparable
HumanEval
GPT-5.2—
Sarvam 30B92.1%
Source
Not directly comparable

Reasoning

ARC-AGI-2
GPT-5.252.9%
Source
Sarvam 30B—
Not directly comparable
ARC-AGI-1
GPT-5.286.17%
Source
Sarvam 30B—
Not directly comparable
GPQA Diamond
GPT-5.2—
Sarvam 30B66.5%
Source
Not directly comparable

Multimodal

MMMU
GPT-5.286.7%
Source
Sarvam 30B—
Not directly comparable
MMMU-Pro
GPT-5.279.5%
Source
Sarvam 30B—
Not directly comparable
RealWorldQA
GPT-5.283.3%
Source
Sarvam 30B—
Not directly comparable
OmniDocBench 1.5
GPT-5.285.7%
Source
Sarvam 30B—
Not directly comparable
Video-MME (with subtitle)
GPT-5.286.0%
Source
Sarvam 30B—
Not directly comparable
Video-MME (w/o subtitle)
GPT-5.285.8%
Source
Sarvam 30B—
Not directly comparable
MathVision
GPT-5.283.0%
Source
Sarvam 30B—
Not directly comparable
We-Math
GPT-5.279.0%
Source
Sarvam 30B—
Not directly comparable
DynaMath
GPT-5.286.8%
Source
Sarvam 30B—
Not directly comparable
SimpleVQA
GPT-5.255.8%
Source
Sarvam 30B—
Not directly comparable
CC-OCR
GPT-5.270.3%
Source
Sarvam 30B—
Not directly comparable
AI2D_TEST
GPT-5.292.2%
Source
Sarvam 30B—
Not directly comparable
CountBench
GPT-5.291.9%
Source
Sarvam 30B—
Not directly comparable
ERQA
GPT-5.259.8%
Source
Sarvam 30B—
Not directly comparable
VideoMMMU
GPT-5.285.9%
Source
Sarvam 30B—
Not directly comparable
MLVU (M-Avg)
GPT-5.285.6%
Source
Sarvam 30B—
Not directly comparable
CharXiv
GPT-5.282.1%
Source
Sarvam 30B—
Not directly comparable
V*
GPT-5.275.9%
Source
Sarvam 30B—
Not directly comparable

Knowledge

GPQA
GPT-5.292.4%
Source
Sarvam 30B—
Not directly comparable
MMLU
GPT-5.2—
Sarvam 30B85.1%
Source
Not directly comparable
MMLU-Pro
GPT-5.2—
Sarvam 30B80%
Source
Not directly comparable

Math

AIME 2025
GPT-5.2100%
Source
Sarvam 30B—
Not directly comparable
FrontierMath v2 (Tiers 1-3)
GPT-5.267.400%
Source
Sarvam 30B—
Not directly comparable
FrontierMath v2 (Tier 4)
GPT-5.231.700%
Source
Sarvam 30B—
Not directly comparable
MATH-500
GPT-5.2—
Sarvam 30B97%
Source
Not directly comparable
HMMT Feb 2025
GPT-5.2—
Sarvam 30B73.3%
Source
Not directly comparable
HMMT Nov 2025
GPT-5.2—
Sarvam 30B74.2%
Source
Not directly comparable

38 public results · 2 shared

Related comparisons

Compare API pricing Read the methodology Open the model selector

Watch GPT-5.2 vs Sarvam 30B

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 5,600+ readers.

Last updated October 17, 2026

GPT-5.2 vs Sarvam 30B

Decision reading

Which one for your work

Long documents

Coding work

Agentic work

Chat turn cost

Cache-heavy agent loop cost

Repository review cost

Which one for a specific job

What is actually comparable

Shape of the matched evidence

Category results, on a stated basis

Knowledge

Instruction following

Agentic

Coding

Reasoning

Multimodal

Multilingual

Math

What each workload costs

Chat turn

Repository review

Cache-heavy agent loop

Specification differences

Context window

API model ID

Cached-input rate

Documented inputs

Documented outputs

Provider availability

Reasoning profile

Weight access

License

Release date

If you already use one of these models

Questions

Which is better, GPT-5.2 or Sarvam 30B?

Which is better for coding, GPT-5.2 or Sarvam 30B?

Which is better for agentic tasks, GPT-5.2 or Sarvam 30B?

Which costs less, GPT-5.2 or Sarvam 30B?

Which has the larger context window, GPT-5.2 or Sarvam 30B?

Self-host vs API cost

Benchmark evidence

Agentic

Coding

Reasoning

Multimodal

Knowledge

Math

Related comparisons

Watch GPT-5.2 vs Sarvam 30B