The Most Spoken Languages of the World

The most-spoken language in the world is English, with around 1.5 billion speakers. The most-spoken native language in the world is Mandarin Chinese, with around 940 million. Most published "top languages" lists pick one of those framings, ignore the other, and ship a ranking that hides the more interesting fact. The interesting fact is the gap.

This article ranks the major languages by both measures, using 2024 Ethnologue / SIL International figures, and walks through what the L1 vs L2 split actually tells you. For the economic-weight angle (which country's GDP each language can transact in), see Languages ranked by share of world GDP. For the cost-to-acquire angle (which language is cheapest for an English speaker to learn), see Easiest languages to learn for English speakers.

Most spoken languages by total speakers (L1 + L2)

Total speakers in 2024, summing native (L1) and second-language (L2) figures from Ethnologue. Numbers are rounded; the underlying counts drift year to year by a few percent and the L2 figures in particular carry meaningful uncertainty.

The top of the table is the headline that propagates across the internet: English first, Mandarin second, Hindi third. That is the L1+L2 ranking. It is correct on its own terms and misleading as a single answer to "which language has the most speakers", because changing one methodological choice (count L1 only) reorders almost everything below the top.

Most spoken languages by L1 (native) speakers only

Strip out the L2 column and the ranking shifts. The languages that lead are the ones whose native-speaker base is demographically large; the lingua-franca giants drop sharply.

Two things to flag on this list. First, English drops from first to third. Second, Western Punjabi enters at ninth, ahead of German. Almost every Western-published "top 10 languages by native speakers" quietly drops Punjabi and substitutes Italian, Turkish, or Vietnamese further down. The figure is not in dispute: around 115 million native Punjabi speakers, concentrated in the Pakistani Punjab and across the border in the Indian Punjab plus the global Punjabi diaspora. It is a top-10 language by any honest L1 count. Its absence from the lists is a Western-publishing artefact, not a data artefact.

What the difference actually tells you

The L1 / L2 ratio for each language tells you what kind of linguistic weight it carries.

English's L2 dominance is the single most important fact in global linguistics. Roughly three times as many people speak English as a second language as speak it natively. No other language on the list has anything like this ratio. The closest comparison is Modern Standard Arabic, where almost no one speaks MSA as a true L1 (everyone grows up with a regional dialect) and the formal MSA register is a learned-at-school register for nearly all of its speakers. English is the world's working language because of the L2 base, not because there are unusually many native English speakers.

Mandarin's near-total reliance on L1 speakers means its global footprint is China-shaped, not lingua-franca-shaped. Outside mainland China, Taiwan, Singapore and the Chinese-speaking diaspora, Mandarin is mostly a heritage language taught at weekend schools to the second-generation children of immigrants. The L2 figure of around 200 million is essentially the population of mainland China that grew up speaking a non-Mandarin Chinese variety (Cantonese, Wu, Hokkien, Hakka and others) and learned Mandarin in school. Mandarin is the language of a country, not the language of the world.

French's L2-heavy ratio is the post-colonial Francophone Africa story in a single number. Roughly 85 million L1 speakers against 225 million L2 speakers means most of the world's French speakers learned it in school, not at home. The L1 base is essentially France plus francophone Belgium plus francophone Switzerland plus Quebec. The L2 base is West and Central Africa, the Maghreb, Madagascar, and a long tail of Indian Ocean and Caribbean francophone territories. The 310 million headline understates how culturally distinct those L2 communities are from Hexagon French; it overstates the homogeneity of the language.

Hindi-Urdu-Punjabi together are a single dialect continuum with roughly 530 million L1 speakers. Counted as Hindustani plus Punjabi, the South Asian Indo-Aryan core comfortably exceeds Spanish on L1 and pushes hard against Mandarin. The decision to count Hindi and Urdu separately is political (different scripts, different official-language histories, different cultural patrons) rather than linguistic (the colloquial spoken forms are largely mutually intelligible). Western lists that count them separately are following the political convention; Indian and Pakistani lists that combine them are following the linguistic reality. Both are defensible. Neither is the only answer.

The languages every list quietly skips

Five languages sit in the top tier by any honest measure and routinely fall off Western-published rankings.

• Punjabi: around 115 million L1 speakers globally, top-10 on any honest L1 list, absent from most "top 10 most spoken" pieces because its speaker base sits in Pakistan and the Indian Punjab and does not fit the European-colonial-language frame the genre defaults to.
• Indonesian / Malay: around 270 million combined speakers across the Malay Archipelago, with an L2-heavy ratio (~80M L1, ~190M L2) that reflects the post-1945 Indonesian-as-national-language project. That project is one of the most successful language-planning interventions in modern history, taking a coastal trade language and turning it into the working language of the world's fourth most populous country in two generations.
• Swahili: around 100 million total speakers across East Africa, predominantly L2, growing African Union official status, and the de facto lingua franca of an economic region that the global rankings still treat as marginal.
• Tagalog / Filipino: around 80 million speakers, the de facto Philippines lingua franca, and structurally significant in Asia-Pacific labour migration.
• Sign languages as a category: there are around 300 distinct sign languages worldwide. American Sign Language (ASL) alone has somewhere between 500,000 and 1 million L1 signers in North America; British Sign Language, French Sign Language, Auslan and Chinese Sign Language are each separate full natural languages with their own grammars. Lists that omit sign languages as a class are wrong on the principle that signed languages are languages.

The pattern these omissions share is geographical, not numerical. Languages whose speaker base sits outside the Atlantic publishing world get rounded down or dropped. Their absence from the headline lists is a Western-publishing artefact, not a reflection of how many people speak them.

Which speaker count matters when you are choosing what to learn

For an English-speaking adult learner deciding on a second language, the speaker-count ranking is a poor decision tool on its own. Four observations make the case.

1. The total-speaker leader (English) does not apply. You already speak it. The marginal speaker count from adding English to your repertoire is zero.
2. The L1 leader (Mandarin) is the most expensive language on the table to acquire. FSI Category V at roughly 2,200 classroom hours to professional working proficiency, against Spanish at around 600-750 hours; the FSI time to fluency calculator tells you what that means for a working adult schedule. Mandarin is a defensible choice if you have specific China-market or family reasons. It is a poor default choice for "I want to talk to a lot of people".
3. The L2-heavy languages come with a built-in community of imperfect speakers. French (225M L2), Indonesian (190M L2), Swahili (mostly L2) and MSA (effectively all L2 above the dialect level) are languages where most speakers are themselves operating in a learned register. That lowers the perfection threshold, lowers the social cost of obvious-foreigner pronunciation, and makes the language genuinely usable at lower CEFR levels than the same hours would deliver in a Mandarin-class L1-dominant language.
4. The cheapest acquisition for English speakers is Spanish. FSI Category I at around 600-750 hours to professional proficiency, around 600 million total speakers across 20-plus countries, the largest learner ecosystem of any second language, and a 42-million-strong US Hispanic market on top. The cost-per-utility ratio is the cleanest on the table; the case is laid out in full at easiest languages to learn for English speakers.

For commercial weight rather than communicative reach, the speaker count is the wrong metric. The GDP-weighted ranking puts English at around 30% of world GDP, Mandarin at 17%, Spanish at 6.5%, and reshuffles the rest of the list in ways the speaker-count ranking does not predict. Use the speaker count to measure reach. Use the GDP table to measure economic surface. Different metric, different answer.

Where the data comes from

The de facto standard source for global speaker counts is Ethnologue, published by SIL International. Their annual catalogue lists around 7,100 living languages with L1 and L2 counts and primary regions for each. The 2024 figures are used throughout this article. The free public version of Ethnologue is now metered; the full database is subscription-only.

Wikipedia's "List of languages by number of speakers" is a useful free aggregator but the underlying numbers are still Ethnologue's plus a few entries from the CIA World Factbook and country-specific censuses. UNESCO publishes the Atlas of the World's Languages in Danger but not a speaker-ranking for major languages. National census data (the Indian decennial census, the US American Community Survey, the UK Census language questions) feed back into the Ethnologue figures.

Two methodological choices matter most.

L2 is defined loosely. Ethnologue's L2 count includes anyone who reports being able to speak the language "well enough" to function in it, which is a low bar in practice. The "1.5 billion English speakers" figure assumes a B1-ish working proficiency at the bottom of the range; tighten the bar to B2 and the figure drops by perhaps a third. Different sources tighten or loosen the bar differently. The order of the top three on the L1+L2 ranking is robust to the choice; the precise numbers are not.

Macro-languages vs individual languages. "Arabic" and "Chinese" are macro-languages with mutually unintelligible variants. Modern Standard Arabic is a learned-at-school register; everyday Arabic is Egyptian, Levantine, Gulf, Maghrebi, Iraqi or Yemeni, and a Moroccan and a Saudi speaking their home varieties to each other are not mutually intelligible without effort. Mandarin, Cantonese, Wu, Hokkien and Hakka are similarly distinct. Different rankings make different choices about whether to count the macro-language or the individual variant, and the result reorders the table. Ethnologue separates the variants; some lists collapse them. This article follows Ethnologue.

A third choice worth flagging: Hindi vs Hindi-Urdu vs Hindustani. The colloquial spoken forms of Hindi and Urdu are largely mutually intelligible; the written registers diverge sharply (Devanagari script and Sanskrit-derived vocabulary for Hindi, Perso-Arabic script and Persian / Arabic-derived vocabulary for Urdu). Lists that count them separately put Hindi at around 610 million total and Urdu at 230 million. Lists that combine them as Hindustani put the combined figure past 800 million and shift the L1 ranking accordingly. Both choices are defensible; both should be flagged when made.

What the trajectory looks like

The L1 rankings will shift over the next 25 years, primarily because of demographics.

Mandarin's L1 base is ageing. China's population peaked in 2022 and is now in net decline. The L1 figure of 940 million will still be the largest by 2050 but the gap to Spanish, Hindi and Bengali will narrow.

Hindi, Bengali and Swahili L1 bases are still growing. Indian and Bangladeshi demographic projections favour both languages climbing the L1 ranking; combined Hindi-Urdu-Punjabi could push past Mandarin on L1 by the late 2040s if the Hindustani grouping is used and current fertility trajectories hold. Swahili is the standout African case: AU official-language status plus East African Community integration plus continued urbanisation in Tanzania and Kenya put it on a trajectory to comfortably exceed 200 million total speakers by 2050.

Spanish L1 is stable to slightly growing. Mexican fertility has dropped sharply but the broader Hispanic American base plus the US Hispanic population keep the L1 figure flat to gently rising.

English L2 keeps climbing, with one open question. The total-speakers figure has grown roughly 50% over the last 25 years on the back of English-medium education in India, Nigeria, the Philippines and the Anglophone tier of Sub-Saharan Africa. The open question is whether conversational AI accelerates that trajectory (cheaper access to high-quality language exposure) or short-circuits it (why learn a second language when your phone can translate). The honest answer is that the second effect is probably overstated by the AI-evangelist commentary and the first effect is probably understated; the L2 base will keep growing for the next decade regardless of what the conversational-AI tooling looks like, because language acquisition is also identity formation and the AI tooling does not substitute for that. Take a sharper position on that question in five years when the data is in.

Cross-links

• Languages ranked by share of world GDP - the economic angle this article deliberately does not cover.
• Easiest languages to learn for English speakers - the cost-per-utility angle, with the FSI Category I list and the case for Spanish.
• Hardest languages to learn - the inverse ranking, with the FSI Category IV / V list and why Mandarin is the expensive bet.
• FSI time to fluency calculator - hours to professional working proficiency for any language, on a working-adult schedule.
• Spanish pillar, French pillar, Mandarin pillar - the per-language hubs.